2025
Bárcena, J. L. Corcuera; Ducange, P.; Marcelloni, F.; Renda, A.
Increasing trust in AI through privacy preservation and model explainability: Federated Learning of Fuzzy Regression Trees Journal Article
In: Information Fusion, vol. 113, 2025.
@article{inffus2025,
title = {Increasing trust in AI through privacy preservation and model explainability: Federated Learning of Fuzzy Regression Trees},
author = {J. L. Corcuera Bárcena and P. Ducange and F. Marcelloni and A. Renda},
url = {https://www.sciencedirect.com/science/article/pii/S1566253524003762},
doi = {10.1016/j.inffus.2024.102598},
year = {2025},
date = {2025-01-01},
journal = {Information Fusion},
volume = {113},
abstract = {Federated Learning (FL) lets multiple data owners collaborate in training a global model without any violation of data privacy, which is a crucial requirement for enhancing users’ trust in Artificial Intelligence (AI) systems. Despite the significant momentum recently gained by the FL paradigm, most of the existing approaches in the field neglect another key pillar for the trustworthiness of AI systems, namely explainability. In this paper, we propose a novel approach for FL of fuzzy regression trees (FRTs), which are generally acknowledged as highly interpretable by-design models. The proposed FL procedure is designed for the scenario of horizontally partitioned data and is based on the transmission of aggregated statistics from the clients to a central server for the tree induction procedure. It is shown that the proposed approach faithfully approximates the ideal case in which the tree induction algorithm is applied on the union of all local datasets, while still ensuring privacy preservation. Furthermore, the FL approach brings benefits, in terms of generalization capability, compared to the local learning setting in which each participant learns its own FRT based only on the private, local, dataset. The adoption of linear models in the leaf nodes ensures a competitive level of performance, as assessed by an extensive experimental campaign on benchmark datasets. The analysis of the results covers both the aspects of accuracy and interpretability of FRT. Finally, we discuss the application of the proposed federated FRT to the task of Quality of Experience forecasting in an automotive case-study.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Federated Learning (FL) lets multiple data owners collaborate in training a global model without any violation of data privacy, which is a crucial requirement for enhancing users’ trust in Artificial Intelligence (AI) systems. Despite the significant momentum recently gained by the FL paradigm, most of the existing approaches in the field neglect another key pillar for the trustworthiness of AI systems, namely explainability. In this paper, we propose a novel approach for FL of fuzzy regression trees (FRTs), which are generally acknowledged as highly interpretable by-design models. The proposed FL procedure is designed for the scenario of horizontally partitioned data and is based on the transmission of aggregated statistics from the clients to a central server for the tree induction procedure. It is shown that the proposed approach faithfully approximates the ideal case in which the tree induction algorithm is applied on the union of all local datasets, while still ensuring privacy preservation. Furthermore, the FL approach brings benefits, in terms of generalization capability, compared to the local learning setting in which each participant learns its own FRT based only on the private, local, dataset. The adoption of linear models in the leaf nodes ensures a competitive level of performance, as assessed by an extensive experimental campaign on benchmark datasets. The analysis of the results covers both the aspects of accuracy and interpretability of FRT. Finally, we discuss the application of the proposed federated FRT to the task of Quality of Experience forecasting in an automotive case-study.
2024
Ducange, Pietro; Fazzolari, Michela; Marcelloni, Francesco
Explainable Intrusion Detection System in IoT Scenarios: A Cross-Device Model Training and Evaluation for Traffic Classification Proceedings Article
In: Proceedings of the 57th Hawaii International Conference on System Sciences, 2024.
@inproceedings{ducange2024explainable,
title = {Explainable Intrusion Detection System in IoT Scenarios: A Cross-Device Model Training and Evaluation for Traffic Classification},
author = {Pietro Ducange and Michela Fazzolari and Francesco Marcelloni},
url = {https://hdl.handle.net/10125/106602},
year = {2024},
date = {2024-01-01},
urldate = {2024-01-01},
booktitle = {Proceedings of the 57th Hawaii International Conference on System Sciences},
abstract = {The proliferation of IoT devices in our daily lives has raised concerns about the security of transmitted data. Due to their limited resources, IoT devices are vulnerable to malware attacks and cyber threats. Detecting and classifying these attacks is crucial to mitigate their impact. Various intrusion detection techniques have been proposed for IoT, including approaches based on ML and AI. Most of the ML/AI-based intrusion detection techniques, though effective, often lack transparency and trustworthiness. To address these aspects, XAI has emerged as a promising solution, providing insights on AI model decisions. In this work, we describe an explainable IDS in IoT networks which embeds a multi-way FDT as an XAI model for traffic classification. We propose a Cross-Device training and evaluation approach in which we evaluate the generalization capability of the IDS when new devices are connected to the IoT network without retraining the FDT.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The proliferation of IoT devices in our daily lives has raised concerns about the security of transmitted data. Due to their limited resources, IoT devices are vulnerable to malware attacks and cyber threats. Detecting and classifying these attacks is crucial to mitigate their impact. Various intrusion detection techniques have been proposed for IoT, including approaches based on ML and AI. Most of the ML/AI-based intrusion detection techniques, though effective, often lack transparency and trustworthiness. To address these aspects, XAI has emerged as a promising solution, providing insights on AI model decisions. In this work, we describe an explainable IDS in IoT networks which embeds a multi-way FDT as an XAI model for traffic classification. We propose a Cross-Device training and evaluation approach in which we evaluate the generalization capability of the IDS when new devices are connected to the IoT network without retraining the FDT.
Ducange, P.; Marcelloni, F.; Renda, A.; Ruffini, F.
Federated Learning of XAI models in healthcare: a case study on Parkinson's Disease Journal Article
In: Cognitive Computation, 2024.
@article{cogcom,
title = {Federated Learning of XAI models in healthcare: a case study on Parkinson's Disease},
author = {P. Ducange and F. Marcelloni and A. Renda and F. Ruffini},
url = {https://link.springer.com/article/10.1007/s12559-024-10332-x},
doi = {10.1007/s12559-024-10332-x},
year = {2024},
date = {2024-01-01},
journal = {Cognitive Computation},
abstract = {Artificial intelligence (AI) systems are increasingly used in healthcare applications, although some challenges have not been completely overcome to make them fully trustworthy and compliant with modern regulations and societal needs. First of all, sensitive health data, essential to train AI systems, are typically stored and managed in several separate medical centers and cannot be shared due to privacy constraints, thus hindering the use of all available information in learning models. Further, transparency and explainability of such systems are becoming increasingly urgent, especially at a time when “opaque” or “black-box” models are commonly used. Recently, technological and algorithmic solutions to these challenges have been investigated: on the one hand, federated learning (FL) has been proposed as a paradigm for collaborative model training among multiple parties without any disclosure of private raw data; on the other hand, research on eXplainable AI (XAI) aims to enhance the explainability of AI systems, either through interpretable by-design approaches or post-hoc explanation techniques. In this paper, we focus on a healthcare case study, namely predicting the progression of Parkinson’s disease, and assume that raw data originate from different medical centers and data collection for centralized training is precluded due to privacy limitations. We aim to investigate how FL of XAI models can allow achieving a good level of accuracy and trustworthiness. Cognitive and biologically inspired approaches are adopted in our analysis: FL of an interpretable by-design fuzzy rule-based system and FL of a neural network explained using a federated version of the SHAP post-hoc explanation technique. We analyze accuracy, interpretability, and explainability of the two approaches, also varying the degree of heterogeneity across several data distribution scenarios. Although the neural network is generally more accurate, the results show that the fuzzy rule-based system achieves competitive performance in the federated setting and presents desirable properties in terms of interpretability and transparency.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Artificial intelligence (AI) systems are increasingly used in healthcare applications, although some challenges have not been completely overcome to make them fully trustworthy and compliant with modern regulations and societal needs. First of all, sensitive health data, essential to train AI systems, are typically stored and managed in several separate medical centers and cannot be shared due to privacy constraints, thus hindering the use of all available information in learning models. Further, transparency and explainability of such systems are becoming increasingly urgent, especially at a time when “opaque” or “black-box” models are commonly used. Recently, technological and algorithmic solutions to these challenges have been investigated: on the one hand, federated learning (FL) has been proposed as a paradigm for collaborative model training among multiple parties without any disclosure of private raw data; on the other hand, research on eXplainable AI (XAI) aims to enhance the explainability of AI systems, either through interpretable by-design approaches or post-hoc explanation techniques. In this paper, we focus on a healthcare case study, namely predicting the progression of Parkinson’s disease, and assume that raw data originate from different medical centers and data collection for centralized training is precluded due to privacy limitations. We aim to investigate how FL of XAI models can allow achieving a good level of accuracy and trustworthiness. Cognitive and biologically inspired approaches are adopted in our analysis: FL of an interpretable by-design fuzzy rule-based system and FL of a neural network explained using a federated version of the SHAP post-hoc explanation technique. We analyze accuracy, interpretability, and explainability of the two approaches, also varying the degree of heterogeneity across several data distribution scenarios. Although the neural network is generally more accurate, the results show that the fuzzy rule-based system achieves competitive performance in the federated setting and presents desirable properties in terms of interpretability and transparency.
Bárcena, J. L. Corcuera; Marcelloni, F.; Renda, A.; Bechini, A.; Ducange, P.
Federated c-means and Fuzzy c-means Clustering Algorithms for Horizontally and Vertically Partitioned Data Journal Article
In: IEEE Transactions on Artificial Intelligence, pp. 1–15, 2024.
@article{tai2024,
title = {Federated c-means and Fuzzy c-means Clustering Algorithms for Horizontally and Vertically Partitioned Data},
author = {J. L. Corcuera Bárcena and F. Marcelloni and A. Renda and A. Bechini and P. Ducange},
url = {https://ieeexplore.ieee.org/document/10595840},
doi = {10.1109/TAI.2024.3426408},
year = {2024},
date = {2024-01-01},
journal = {IEEE Transactions on Artificial Intelligence},
pages = {1–15},
abstract = {Federated clustering lets multiple data owners collaborate in discovering patterns from distributed data without violating privacy requirements. The federated versions of traditional clustering algorithms proposed so far are, however, “lossy” since they fail to identify exactly the same clusters as the original versions executed on the merged data stored in a centralized server, as would happen if no privacy constraint occurred. In this paper, we propose federated procedures for losslessly executing the C-Means (CM) and the Fuzzy C-Means (FCM) algorithms in both horizontally and vertically partitioned data scenarios, while preserving data privacy. We formally prove that the proposed federated procedures identify the same clusters determined by applying the algorithms to the union of all local data. Further, we present an extensive experimental analysis for characterizing the behavior of the proposed approach in a typical federated learning scenario, that is, as the fraction of participants in the federation changes. We focus on the federated FCM and the horizontally partitioned data, which is the most interesting scenario. We show that the proposed procedure is effective and is able to achieve competitive performance with respect to two recently proposed versions of federated FCM for horizontally partitioned data.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Federated clustering lets multiple data owners collaborate in discovering patterns from distributed data without violating privacy requirements. The federated versions of traditional clustering algorithms proposed so far are, however, “lossy” since they fail to identify exactly the same clusters as the original versions executed on the merged data stored in a centralized server, as would happen if no privacy constraint occurred. In this paper, we propose federated procedures for losslessly executing the C-Means (CM) and the Fuzzy C-Means (FCM) algorithms in both horizontally and vertically partitioned data scenarios, while preserving data privacy. We formally prove that the proposed federated procedures identify the same clusters determined by applying the algorithms to the union of all local data. Further, we present an extensive experimental analysis for characterizing the behavior of the proposed approach in a typical federated learning scenario, that is, as the fraction of participants in the federation changes. We focus on the federated FCM and the horizontally partitioned data, which is the most interesting scenario. We show that the proposed procedure is effective and is able to achieve competitive performance with respect to two recently proposed versions of federated FCM for horizontally partitioned data.
Ducange, P.; Marcelloni, F.; Renda, A.; Ruffini, F.
Consistent Post-Hoc Explainability in Federated Learning through Federated Fuzzy Clustering Proceedings Article
In: IEEE International Conference on Fuzzy Systems, 2024.
@inproceedings{fuzzieee2024_shap,
title = {Consistent Post-Hoc Explainability in Federated Learning through Federated Fuzzy Clustering},
author = {P. Ducange and F. Marcelloni and A. Renda and F. Ruffini},
url = {https://ieeexplore.ieee.org/document/10611761},
doi = {10.1109/FUZZ-IEEE60900.2024.10611761},
year = {2024},
date = {2024-01-01},
booktitle = {IEEE International Conference on Fuzzy Systems},
abstract = {Ensuring trustworthiness of AI systems by enforcing, for instance, data privacy and model explainability, has become urgent in our society. Recently, the Federated Learning (FL) paradigm has been proposed to preserve data privacy during collaborative model learning. Unfortunately, FL poses critical challenges in the application of post-hoc explanation methods which are used to explain opaque models such as neural networks. In this paper we present an approach for enhancing the explainability of opaque models generated according to the FL paradigm. We focus on one of the most popular methods, namely SHapley Additive exPlanations method (SHAP). Given an input instance, SHAP can explain why an opaque model generated that specific output prediction from the input values. To provide the explanation SHAP needs access to a background dataset, typically consisting of representative training instances. In FL setting, however, the training data are scattered over multiple participants and cannot be shared due to privacy constraints. On the other side, the background dataset should be representative of the overall training set. To this aim, we propose to adopt a federated Fuzzy C-Means clustering for the generation of a common background dataset made up of cluster centers. The resulting background dataset is representative of the actual distribution of the data and can be made available to all participants without violating privacy, thus ensuring accuracy and consistency of the explanations. A thorough experimental analysis shows the validity of the proposed approach also in comparison with baseline and alternative approaches.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Ensuring trustworthiness of AI systems by enforcing, for instance, data privacy and model explainability, has become urgent in our society. Recently, the Federated Learning (FL) paradigm has been proposed to preserve data privacy during collaborative model learning. Unfortunately, FL poses critical challenges in the application of post-hoc explanation methods which are used to explain opaque models such as neural networks. In this paper we present an approach for enhancing the explainability of opaque models generated according to the FL paradigm. We focus on one of the most popular methods, namely SHapley Additive exPlanations method (SHAP). Given an input instance, SHAP can explain why an opaque model generated that specific output prediction from the input values. To provide the explanation SHAP needs access to a background dataset, typically consisting of representative training instances. In FL setting, however, the training data are scattered over multiple participants and cannot be shared due to privacy constraints. On the other side, the background dataset should be representative of the overall training set. To this aim, we propose to adopt a federated Fuzzy C-Means clustering for the generation of a common background dataset made up of cluster centers. The resulting background dataset is representative of the actual distribution of the data and can be made available to all participants without violating privacy, thus ensuring accuracy and consistency of the explanations. A thorough experimental analysis shows the validity of the proposed approach also in comparison with baseline and alternative approaches.
Daole, M.; Ducange, P.; Marcelloni, F.; Renda, A.
Trustworthy AI in Heterogeneous Settings: Federated Learning of Explainable Classifiers Proceedings Article
In: IEEE International Conference on Fuzzy Systems, 2024.
@inproceedings{fuzzieee2024_chi,
title = {Trustworthy AI in Heterogeneous Settings: Federated Learning of Explainable Classifiers},
author = {M. Daole and P. Ducange and F. Marcelloni and A. Renda},
url = {https://ieeexplore.ieee.org/document/10612109},
doi = {10.1109/FUZZ-IEEE60900.2024.10612109},
year = {2024},
date = {2024-01-01},
booktitle = {IEEE International Conference on Fuzzy Systems},
abstract = {Trustworthy Artificial Intelligence (AI) has gained significant relevance worldwide. Federated Learning (FL) and eXplainable Artificial Intelligence (XAI) are two among the most relevant paradigms for accomplishing the requirements of trustworthy AI-based applications. On the one hand, FL guarantees data privacy throughout a collaborative learning of an AI model from decentralized data. On the other hand, XAI models ensure transparency, accountability, and trust in AI-based systems by providing understandable explanations for their predictions and decisions. To the best of our knowledge, only few works have explored the combination of FL with inherently explainable models, especially for classification task. In this work, we investigate FL of explainable classifiers, namely Fuzzy Rule-based Classifiers. In the proposed FL scheme, each participant creates its own set of classification rules from its own local training data, resorting to a simple procedure that generates a rule for each training instance. Local rules are sent to a central server which is in charge of aggregating them by removing duplicates and solving conflicts. The aggregated set of rules is then forwarded to the single participants for inference purposes. In our experimental analysis we consider two real-world case studies focusing on heterogeneous settings, namely non-IID (Independent and Identically Distributed) scenarios. Our FL scheme offers significant advantages in terms of classification performance to the participants in the federation, preserving data privacy.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Trustworthy Artificial Intelligence (AI) has gained significant relevance worldwide. Federated Learning (FL) and eXplainable Artificial Intelligence (XAI) are two among the most relevant paradigms for accomplishing the requirements of trustworthy AI-based applications. On the one hand, FL guarantees data privacy throughout a collaborative learning of an AI model from decentralized data. On the other hand, XAI models ensure transparency, accountability, and trust in AI-based systems by providing understandable explanations for their predictions and decisions. To the best of our knowledge, only few works have explored the combination of FL with inherently explainable models, especially for classification task. In this work, we investigate FL of explainable classifiers, namely Fuzzy Rule-based Classifiers. In the proposed FL scheme, each participant creates its own set of classification rules from its own local training data, resorting to a simple procedure that generates a rule for each training instance. Local rules are sent to a central server which is in charge of aggregating them by removing duplicates and solving conflicts. The aggregated set of rules is then forwarded to the single participants for inference purposes. In our experimental analysis we consider two real-world case studies focusing on heterogeneous settings, namely non-IID (Independent and Identically Distributed) scenarios. Our FL scheme offers significant advantages in terms of classification performance to the participants in the federation, preserving data privacy.
Basile, Miriam; Gibiino, Fabio; Cavazza, Jacopo; Semplici, Paolo; Cocco, Martina; Marcelloni, Francesco; Bechini, Alessio; Vanello, Nicola
Unsupervised Learning of Speckle Removal from Real Ultrasound Acquisitions without Clean Data Proceedings Article
In: 2024 IEEE International Symposium on Medical Measurements and Applications (MeMeA), pp. 1-6, 2024.
@inproceedings{10596830,
title = {Unsupervised Learning of Speckle Removal from Real Ultrasound Acquisitions without Clean Data},
author = {Miriam Basile and Fabio Gibiino and Jacopo Cavazza and Paolo Semplici and Martina Cocco and Francesco Marcelloni and Alessio Bechini and Nicola Vanello},
doi = {10.1109/MeMeA60663.2024.10596830},
year = {2024},
date = {2024-01-01},
booktitle = {2024 IEEE International Symposium on Medical Measurements and Applications (MeMeA)},
pages = {1-6},
abstract = {Ultrasound (US) images suffer from speckle noise, a granular pattern that hampers contrast and resolution, making low-contrast structures critically difficult to identify. Albeit traditional filtering and machine learning approaches can handle this problem, both have limitations: such as the need of (hyper-)parameters fine-tuning or the necessity of data collection and annotation. In our study, we explored an unsupervised image filtering method based on blind denoising, so that we can systematically overcome the need of ground truth annotations. Our approach is based on a noise2noise u-net backbone (N2N) fed by a novel image representation approach. Dubbed Emulated Frequency Compound (EFQ), this study is intended to propose and validate it in the small data regime which is compatible with the typical applicative scenario of US imaging. As our experimental validation shows, the adoption of EFQ for N2N results in a favorable performance with respect to a number of state-of-the-art methods and related baselines.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Ultrasound (US) images suffer from speckle noise, a granular pattern that hampers contrast and resolution, making low-contrast structures critically difficult to identify. Albeit traditional filtering and machine learning approaches can handle this problem, both have limitations: such as the need of (hyper-)parameters fine-tuning or the necessity of data collection and annotation. In our study, we explored an unsupervised image filtering method based on blind denoising, so that we can systematically overcome the need of ground truth annotations. Our approach is based on a noise2noise u-net backbone (N2N) fed by a novel image representation approach. Dubbed Emulated Frequency Compound (EFQ), this study is intended to propose and validate it in the small data regime which is compatible with the typical applicative scenario of US imaging. As our experimental validation shows, the adoption of EFQ for N2N results in a favorable performance with respect to a number of state-of-the-art methods and related baselines.
Bondielli, Alessandro; Dell'Oglio, Pietro; Lenci, Alessandro; Marcelloni, Francesco; Passaro, Lucia
Dataset for multimodal fake news detection and verification tasks Journal Article
In: Data in Brief, vol. 54, pp. 110440, 2024, ISSN: 2352-3409.
@article{BONDIELLI2024110440,
title = {Dataset for multimodal fake news detection and verification tasks},
author = {Alessandro Bondielli and Pietro Dell'Oglio and Alessandro Lenci and Francesco Marcelloni and Lucia Passaro},
url = {https://www.sciencedirect.com/science/article/pii/S2352340924004098},
doi = {https://doi.org/10.1016/j.dib.2024.110440},
issn = {2352-3409},
year = {2024},
date = {2024-01-01},
journal = {Data in Brief},
volume = {54},
pages = {110440},
abstract = {The proliferation of online disinformation and fake news, particularly in the context of breaking news events, demands the development of effective detection mechanisms. While textual content remains the predominant medium for disseminating misleading information, the contribution of other modalities is increasingly emerging within online outlets and social media platforms. However, multimodal datasets, which incorporate diverse modalities such as texts and images, are not very common yet, especially in low-resource languages. This study addresses this gap by releasing a dataset tailored for multimodal fake news detection in the Italian language. This dataset was originally employed in a shared task on the Italian language. The dataset is divided into two data subsets, each corresponding to a distinct sub-task. In sub-task 1, the goal is to assess the effectiveness of multimodal fake news detection systems. Sub-task 2 aims to delve into the interplay between text and images, specifically analyzing how these modalities mutually influence the interpretation of content when distinguishing between fake and real news. Both sub-tasks were managed as classification problems. The dataset consists of social media posts and news articles. After collecting it, it was labeled via crowdsourcing. Annotators were provided with external knowledge about the topic of the news to be labeled, enhancing their ability to discriminate between fake and real news. The data subsets for sub-task 1 and sub-task 2 consist of 913 and 1350 items, respectively, encompassing newspaper articles and tweets.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The proliferation of online disinformation and fake news, particularly in the context of breaking news events, demands the development of effective detection mechanisms. While textual content remains the predominant medium for disseminating misleading information, the contribution of other modalities is increasingly emerging within online outlets and social media platforms. However, multimodal datasets, which incorporate diverse modalities such as texts and images, are not very common yet, especially in low-resource languages. This study addresses this gap by releasing a dataset tailored for multimodal fake news detection in the Italian language. This dataset was originally employed in a shared task on the Italian language. The dataset is divided into two data subsets, each corresponding to a distinct sub-task. In sub-task 1, the goal is to assess the effectiveness of multimodal fake news detection systems. Sub-task 2 aims to delve into the interplay between text and images, specifically analyzing how these modalities mutually influence the interpretation of content when distinguishing between fake and real news. Both sub-tasks were managed as classification problems. The dataset consists of social media posts and news articles. After collecting it, it was labeled via crowdsourcing. Annotators were provided with external knowledge about the topic of the news to be labeled, enhancing their ability to discriminate between fake and real news. The data subsets for sub-task 1 and sub-task 2 consist of 913 and 1350 items, respectively, encompassing newspaper articles and tweets.
Ducange, Pietro; Fazzolari, Michela; Marcelloni, Francesco; Marino, Martina; Matrella, Roberta
Continuous Monitoring of Body Shaming Actions in Social Networks Proceedings Article
In: 2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS), pp. 1-8, 2024.
@inproceedings{10570011,
title = {Continuous Monitoring of Body Shaming Actions in Social Networks},
author = {Pietro Ducange and Michela Fazzolari and Francesco Marcelloni and Martina Marino and Roberta Matrella},
doi = {10.1109/EAIS58494.2024.10570011},
year = {2024},
date = {2024-01-01},
booktitle = {2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS)},
pages = {1-8},
abstract = {Social media platforms provide significant communication opportunities, yet they are sometimes utilized inappropriately. Among the prevalent misuses is cyberbullying, particularly body shaming. Twitter (now known as X) facilitates daily interactions for millions of users, allowing them to post and read succinct messages on the Internet. This study explores the development of a text mining pipeline, employing machine learning models, designed to identify instances of body shaming within the Italian Twitter community. Additionally, it addresses the challenge of concept drift, a phenomenon where the characteristics of the dataset evolve over time, potentially leading to a decline in classification accuracy. Through an online monitoring phase, the presence of concept drift is confirmed, and an effective solution is sought by evaluating several strategies to mitigate its impact.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Social media platforms provide significant communication opportunities, yet they are sometimes utilized inappropriately. Among the prevalent misuses is cyberbullying, particularly body shaming. Twitter (now known as X) facilitates daily interactions for millions of users, allowing them to post and read succinct messages on the Internet. This study explores the development of a text mining pipeline, employing machine learning models, designed to identify instances of body shaming within the Italian Twitter community. Additionally, it addresses the challenge of concept drift, a phenomenon where the characteristics of the dataset evolve over time, potentially leading to a decline in classification accuracy. Through an online monitoring phase, the presence of concept drift is confirmed, and an effective solution is sought by evaluating several strategies to mitigate its impact.
Dell'Oglio, Pietro; Bondielli, Alessandro; Marcelloni, Francesco
A System for Assisting Users in Automatically Obtaining Comprehensive and Condensed Information About an Event from Various Sources Proceedings Article
In: Abraham, Ajith; Bajaj, Anu; Hanne, Thomas; Siarry, Patrick; Ma, Kun (Ed.): Intelligent Systems Design and Applications, pp. 483–492, Springer Nature Switzerland, Cham, 2024, ISBN: 978-3-031-64847-2.
@inproceedings{10.1007/978-3-031-64847-2_45,
title = {A System for Assisting Users in Automatically Obtaining Comprehensive and Condensed Information About an Event from Various Sources},
author = {Pietro Dell'Oglio and Alessandro Bondielli and Francesco Marcelloni},
editor = {Ajith Abraham and Anu Bajaj and Thomas Hanne and Patrick Siarry and Kun Ma},
isbn = {978-3-031-64847-2},
year = {2024},
date = {2024-01-01},
booktitle = {Intelligent Systems Design and Applications},
pages = {483–492},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {In today's world, the majority of newspapers utilize social media platforms to distribute their published information. Thus, users have available a huge amount of news on the same topic. Nevertheless, they prefer to avoid the arduous task of reading every single news article for forming their independent own opinion. To address this challenge, this paper proposes a system that exploits transformer neural models to discern the differences in content between an article read by a user and a collection of related articles about the same topic, and text summarization models to summarize these differences. Thus, the user can read only the summary rather than all the articles still acquiring the different opinions on the topic from the different sources without the burden of reading all the related articles. The system has been tested against human opinion collected through crowdsourcing, obtaining an average F1-Score of 0.753, and compared with a BERT classifier which achieved similar accuracy with an average F1-score of 0.754. In our discussion, we delve into the advantages and critical issues that emerged during the evaluation, along with potential future directions.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In today's world, the majority of newspapers utilize social media platforms to distribute their published information. Thus, users have available a huge amount of news on the same topic. Nevertheless, they prefer to avoid the arduous task of reading every single news article for forming their independent own opinion. To address this challenge, this paper proposes a system that exploits transformer neural models to discern the differences in content between an article read by a user and a collection of related articles about the same topic, and text summarization models to summarize these differences. Thus, the user can read only the summary rather than all the articles still acquiring the different opinions on the topic from the different sources without the burden of reading all the related articles. The system has been tested against human opinion collected through crowdsourcing, obtaining an average F1-Score of 0.753, and compared with a BERT classifier which achieved similar accuracy with an average F1-score of 0.754. In our discussion, we delve into the advantages and critical issues that emerged during the evaluation, along with potential future directions.
Pecori, Riccardo; Panella, Giovanni; Vurro, Filippo; Bettelli, Manuele; Fazzolari, Michela; Ducange, Pietro
An Explainable Smart Agriculture System based on In- Vivo Biosensors Proceedings Article
In: 2024 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 1-8, 2024.
@inproceedings{10612040,
title = {An Explainable Smart Agriculture System based on In- Vivo Biosensors},
author = {Riccardo Pecori and Giovanni Panella and Filippo Vurro and Manuele Bettelli and Michela Fazzolari and Pietro Ducange},
doi = {10.1109/FUZZ-IEEE60900.2024.10612040},
year = {2024},
date = {2024-01-01},
booktitle = {2024 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE)},
pages = {1-8},
abstract = {Some of the most significant factors regarding plant growth and food production are for sure water stress and drought. Predicting the water stress of crops in advance with respect to its visible signs is priceless and could permit one to intervene early to restore healthy growth conditions. In this paper, we discuss an Explainable Smart Agriculture System for monitoring the water stress status of tomato plants based on a novel in-vivo biosensor. Specifically, we embed, in the proposed system, an intrinsically explainable classifier, namely a fuzzy decision tree, to characterize the status of the plants in four different categories. To this aim, we extract four features related to the ionic currents inside the sap of the plants themselves. Thanks to the explainable classifier, we offer insights into the classification of the status of the plants. This contributes to a deeper understanding of the unseen processes occurring within the plants, enabling early detection of stress due to water shortage before it becomes visibly apparent. We evaluate the effectiveness of our approach considering the real data extracted from in-vivo biosensors deployed on two different types of tomato plants. Preliminary results show that the proposed explainable classifier achieves promising results in terms of both explainability and classification capability. Additionally, we present and discuss some examples of rules derived from the decision trees, emphasizing their significance in understanding the sap activities within plants. This under-standing aids in implementing effective countermeasures, for example in real-world on-the-field automated irrigation systems, to maintain plant health.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Some of the most significant factors regarding plant growth and food production are for sure water stress and drought. Predicting the water stress of crops in advance with respect to its visible signs is priceless and could permit one to intervene early to restore healthy growth conditions. In this paper, we discuss an Explainable Smart Agriculture System for monitoring the water stress status of tomato plants based on a novel in-vivo biosensor. Specifically, we embed, in the proposed system, an intrinsically explainable classifier, namely a fuzzy decision tree, to characterize the status of the plants in four different categories. To this aim, we extract four features related to the ionic currents inside the sap of the plants themselves. Thanks to the explainable classifier, we offer insights into the classification of the status of the plants. This contributes to a deeper understanding of the unseen processes occurring within the plants, enabling early detection of stress due to water shortage before it becomes visibly apparent. We evaluate the effectiveness of our approach considering the real data extracted from in-vivo biosensors deployed on two different types of tomato plants. Preliminary results show that the proposed explainable classifier achieves promising results in terms of both explainability and classification capability. Additionally, we present and discuss some examples of rules derived from the decision trees, emphasizing their significance in understanding the sap activities within plants. This under-standing aids in implementing effective countermeasures, for example in real-world on-the-field automated irrigation systems, to maintain plant health.
Duarte-Martínez, V.; Perez, I. J.; Ducange, P.; Cobo, M. J
Understanding the Conceptual Structure of Large Language Models through Bibliographical Network Proceedings Article
In: 2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS), pp. 1-7, 2024.
@inproceedings{10569112,
title = {Understanding the Conceptual Structure of Large Language Models through Bibliographical Network},
author = {V. Duarte-Martínez and I. J. Perez and P. Ducange and M. J Cobo},
doi = {10.1109/EAIS58494.2024.10569112},
year = {2024},
date = {2024-01-01},
booktitle = {2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS)},
pages = {1-7},
abstract = {Large Language Models represent a transformative technology at the forefront of artificial intelligence and natural language processing, with applications spanning diverse domains. This study conducts a comprehensive science mapping analysis of the LLMs research field, leveraging bibliometric techniques to uncover its thematic structure, trends, and global actors involved. Utilizing data from the Web of Science, a corpus of 1303 research documents from 2010 to 2023 is analyzed, revealing a notable surge in research activity, particularly in recent years. Key thematic areas driving research within the field are identified, including chatbot, code generation, augmented reality, transformers, and machine learning paradigms. Foundational technologies such as transformers are pivotal in shaping the research landscape, while emerging themes like prompt learning hint at future directions. This study offers valuable insights for researchers, practitioners, and policymakers seeking to navigate the dynamic landscape of LLMs research and harness its full potential for societal benefit.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Large Language Models represent a transformative technology at the forefront of artificial intelligence and natural language processing, with applications spanning diverse domains. This study conducts a comprehensive science mapping analysis of the LLMs research field, leveraging bibliometric techniques to uncover its thematic structure, trends, and global actors involved. Utilizing data from the Web of Science, a corpus of 1303 research documents from 2010 to 2023 is analyzed, revealing a notable surge in research activity, particularly in recent years. Key thematic areas driving research within the field are identified, including chatbot, code generation, augmented reality, transformers, and machine learning paradigms. Foundational technologies such as transformers are pivotal in shaping the research landscape, while emerging themes like prompt learning hint at future directions. This study offers valuable insights for researchers, practitioners, and policymakers seeking to navigate the dynamic landscape of LLMs research and harness its full potential for societal benefit.
Panella, Giovanni; Ducange, Pietro; Bettelli, Manuele; Vurro, Filippo; Fazzolari, Michela; Pecori, Riccardo
Leveraging Incremental Decision Trees and In-Vivo Biosensors for an Explainable Plant Health Monitoring System Proceedings Article
In: 2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS), pp. 1-8, 2024.
@inproceedings{10570002,
title = {Leveraging Incremental Decision Trees and In-Vivo Biosensors for an Explainable Plant Health Monitoring System},
author = {Giovanni Panella and Pietro Ducange and Manuele Bettelli and Filippo Vurro and Michela Fazzolari and Riccardo Pecori},
doi = {10.1109/EAIS58494.2024.10570002},
year = {2024},
date = {2024-01-01},
booktitle = {2024 IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS)},
pages = {1-8},
abstract = {Among the factors concerning plant development and agricultural yield, water stress and drought emerge as pivotal factors. Indeed, the ability to know in advance imminent water stress in crops based on measurable biochemical metrics is priceless, as it offers the opportunity for rapid interventions aimed at restoring optimal growth conditions before the plants show clear visible stress symptoms.In this work, we present an explainable system for smart agriculture focused on the continuous monitoring of the water stress condition of tomato plants, achieved through a new in-vivo biosensor, named bioristor. The proposed system embeds an incremental and explainable by design classifier. Specifically, we experimented with the traditional Hoeffding decision tree and its fuzzy version. This system analyzes the data received from bioristors to assess the health status of a tomato plant and classifies it into four classes. The proposed system also leverages an incremental learning technique, which allows the classification model to be updated during the monitoring period, to maintain adequate classification performance. In this way, the conditions of the plants are monitored continuously with an effective model, allowing for timely countermeasures to be taken if a water stress situation is detected. We present preliminary results on a real dataset, using four features related to the ionic currents within the plant sap, measured through bioristors. We assessed the system performance both in terms of classification ability and model complexity, obtaining promising results and the generation of interesting rules that could allow the implementation of effective countermeasures to keep the plants healthy as long as possible.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Among the factors concerning plant development and agricultural yield, water stress and drought emerge as pivotal factors. Indeed, the ability to know in advance imminent water stress in crops based on measurable biochemical metrics is priceless, as it offers the opportunity for rapid interventions aimed at restoring optimal growth conditions before the plants show clear visible stress symptoms.In this work, we present an explainable system for smart agriculture focused on the continuous monitoring of the water stress condition of tomato plants, achieved through a new in-vivo biosensor, named bioristor. The proposed system embeds an incremental and explainable by design classifier. Specifically, we experimented with the traditional Hoeffding decision tree and its fuzzy version. This system analyzes the data received from bioristors to assess the health status of a tomato plant and classifies it into four classes. The proposed system also leverages an incremental learning technique, which allows the classification model to be updated during the monitoring period, to maintain adequate classification performance. In this way, the conditions of the plants are monitored continuously with an effective model, allowing for timely countermeasures to be taken if a water stress situation is detected. We present preliminary results on a real dataset, using four features related to the ionic currents within the plant sap, measured through bioristors. We assessed the system performance both in terms of classification ability and model complexity, obtaining promising results and the generation of interesting rules that could allow the implementation of effective countermeasures to keep the plants healthy as long as possible.
Olivelli, Martina; Donati, Massimiliano; Vianello, Annamaria; Petrucci, Ilaria; Masi, Stefano; Bechini, Alessio; Fanucci, Luca
Enhancing Precision of Telemonitoring of COVID-19 Patients through Expert System Based on IoT Data Elaboration Journal Article
In: Electronics, vol. 13, no. 8, 2024, ISSN: 2079-9292.
@article{electronics13081462,
title = {Enhancing Precision of Telemonitoring of COVID-19 Patients through Expert System Based on IoT Data Elaboration},
author = {Martina Olivelli and Massimiliano Donati and Annamaria Vianello and Ilaria Petrucci and Stefano Masi and Alessio Bechini and Luca Fanucci},
url = {https://www.mdpi.com/2079-9292/13/8/1462},
doi = {10.3390/electronics13081462},
issn = {2079-9292},
year = {2024},
date = {2024-01-01},
journal = {Electronics},
volume = {13},
number = {8},
abstract = {The emergence of the highly contagious coronavirus disease has led to multiple pandemic waves, resulting in a significant number of hospitalizations and fatalities. Even outside of hospitals, general practitioners have faced serious challenges, stretching their resources and putting themselves at risk of infection. Telemonitoring systems based on Internet of things technology have emerged as valuable tools for remotely monitoring disease progression, facilitating rapid intervention, and reducing the risk of hospitalization and mortality. They allow for personalized monitoring strategies and tailored treatment plans, which are crucial for improving health outcomes. However, determining the appropriate monitoring intensity remains the responsibility of physicians, which poses challenges and impacts their workload, and thus, can hinder timely responses. To address these challenges, this paper proposes an expert system designed to recommend and adjust the monitoring intensity for COVID-19 patients receiving home treatment based on their medical history, vital signs, and reported symptoms. The system underwent initial validation using real-world cases, demonstrating a favorable performance (F1-score of 0.85). Subsequently, once integrated with an Internet of Things telemonitoring system, a clinical trial will assess the system’s reliability in creating telemonitoring plans comparable with those of medics, evaluate its effectiveness in reducing medic–patient interactions or hospitalizations, and gauge patient satisfaction and safety.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The emergence of the highly contagious coronavirus disease has led to multiple pandemic waves, resulting in a significant number of hospitalizations and fatalities. Even outside of hospitals, general practitioners have faced serious challenges, stretching their resources and putting themselves at risk of infection. Telemonitoring systems based on Internet of things technology have emerged as valuable tools for remotely monitoring disease progression, facilitating rapid intervention, and reducing the risk of hospitalization and mortality. They allow for personalized monitoring strategies and tailored treatment plans, which are crucial for improving health outcomes. However, determining the appropriate monitoring intensity remains the responsibility of physicians, which poses challenges and impacts their workload, and thus, can hinder timely responses. To address these challenges, this paper proposes an expert system designed to recommend and adjust the monitoring intensity for COVID-19 patients receiving home treatment based on their medical history, vital signs, and reported symptoms. The system underwent initial validation using real-world cases, demonstrating a favorable performance (F1-score of 0.85). Subsequently, once integrated with an Internet of Things telemonitoring system, a clinical trial will assess the system’s reliability in creating telemonitoring plans comparable with those of medics, evaluate its effectiveness in reducing medic–patient interactions or hospitalizations, and gauge patient satisfaction and safety.
Schiavo, Alessio; Marcelloni, Francesco; Ducange, Pietro; Renda, Alessandro; Ruffini, Fabrizio; Dani, Renzo
On Predicting Spare Parts for Field Services by Leveraging Fault Description and Historical Repairing Data Journal Article
In: IEEE Access, pp. 1-1, 2024.
@article{10737069,
title = {On Predicting Spare Parts for Field Services by Leveraging Fault Description and Historical Repairing Data},
author = {Alessio Schiavo and Francesco Marcelloni and Pietro Ducange and Alessandro Renda and Fabrizio Ruffini and Renzo Dani},
url = {https://ieeexplore.ieee.org/document/10737069},
doi = {10.1109/ACCESS.2024.3487493},
year = {2024},
date = {2024-01-01},
journal = {IEEE Access},
pages = {1-1},
abstract = {The traditional process of repairing electrical appliances, including industrial appliances, is activated by a fault report submitted by the owner of the appliance that summarizes the symptoms of the fault. Then, based on their experience and on the fault report, the field service technicians visit the customer, bringing the spare parts they consider necessary. If the on-site repair requires spare parts that the technicians did not bring, they are forced to return to the customer’s location, thus increasing costs and causing customer dissatisfaction. To improve the success rate and consequently to reduce the number of visits, we propose a Spare Parts Prediction (SPP) system which integrates principles from Case-Based Reasoning and Natural Language Processing. This system recommends a sorted list of spare parts by analyzing the appliance fault descriptions and leveraging historical data of repair and maintenance interventions. Application of the SPP system in a real-world scenario presented a substantial reduction of multiple visits up to 50%.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The traditional process of repairing electrical appliances, including industrial appliances, is activated by a fault report submitted by the owner of the appliance that summarizes the symptoms of the fault. Then, based on their experience and on the fault report, the field service technicians visit the customer, bringing the spare parts they consider necessary. If the on-site repair requires spare parts that the technicians did not bring, they are forced to return to the customer’s location, thus increasing costs and causing customer dissatisfaction. To improve the success rate and consequently to reduce the number of visits, we propose a Spare Parts Prediction (SPP) system which integrates principles from Case-Based Reasoning and Natural Language Processing. This system recommends a sorted list of spare parts by analyzing the appliance fault descriptions and leveraging historical data of repair and maintenance interventions. Application of the SPP system in a real-world scenario presented a substantial reduction of multiple visits up to 50%.
2023
Merluzzi, Mattia; Borsos, Tamás; Rajatheva, Nandana; Benczúr, András A.; Farhadi, Hamed; Yassine, Taha; Mück, Markus Dominik; Barmpounakis, Sokratis; Strinati, Emilio Calvanese; Dampahalage, Dilin; Demestichas, Panagiotis; Ducange, Pietro; Filippou, Miltiadis C.; Baltar, Leonardo Gomes; Haraldson, Johan; Karaçay, Leyli; Korpi, Dani; Lamprousi, Vasiliki; Marcelloni, Francesco; Mohammadi, Jafar; Rajapaksha, Nuwanthika; Renda, Alessandro; Uusitalo, Mikko A.
The Hexa-X project vision on Artificial Intelligence and Machine Learning-driven Communication and Computation co-design for 6G Journal Article
In: IEEE Access, pp. 1-1, 2023.
@article{10156818,
title = {The Hexa-X project vision on Artificial Intelligence and Machine Learning-driven Communication and Computation co-design for 6G},
author = {Mattia Merluzzi and Tamás Borsos and Nandana Rajatheva and András A. Benczúr and Hamed Farhadi and Taha Yassine and Markus Dominik Mück and Sokratis Barmpounakis and Emilio Calvanese Strinati and Dilin Dampahalage and Panagiotis Demestichas and Pietro Ducange and Miltiadis C. Filippou and Leonardo Gomes Baltar and Johan Haraldson and Leyli Karaçay and Dani Korpi and Vasiliki Lamprousi and Francesco Marcelloni and Jafar Mohammadi and Nuwanthika Rajapaksha and Alessandro Renda and Mikko A. Uusitalo},
doi = {10.1109/ACCESS.2023.3287939},
year = {2023},
date = {2023-01-01},
journal = {IEEE Access},
pages = {1-1},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Cavalieri, A.; Ducange, Pietro; Fabi, S.; Russo, F.; Tonellotto, Nicola
An Intelligent system for the categorization of question time official documents of the Italian Chamber of Deputies Journal Article
In: Journal of Information Technology & Politics, vol. 20, no. 3, pp. 213-234, 2023.
@article{doi:10.1080/19331681.2022.2082622,
title = {An Intelligent system for the categorization of question time official documents of the Italian Chamber of Deputies},
author = {A. Cavalieri and Pietro Ducange and S. Fabi and F. Russo and Nicola Tonellotto},
url = {https://doi.org/10.1080/19331681.2022.2082622},
doi = {10.1080/19331681.2022.2082622},
year = {2023},
date = {2023-01-01},
journal = {Journal of Information Technology & Politics},
volume = {20},
number = {3},
pages = {213-234},
publisher = {Routledge},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Bechini, Alessio; Bondielli, Alessandro; Dell'Oglio, Pietro; Marcelloni, Francesco
From basic approaches to novel challenges and applications in Sequential Pattern Mining Journal Article
In: Applied Computing and Intelligence, vol. 3, no. 1, pp. 44-78, 2023, ISSN: 2771-392X.
@article{nokey,
title = {From basic approaches to novel challenges and applications in Sequential Pattern Mining},
author = {Alessio Bechini and Alessandro Bondielli and Pietro Dell'Oglio and Francesco Marcelloni},
url = {https://www.aimspress.com/article/doi/10.3934/aci.2023004},
doi = {10.3934/aci.2023004},
issn = {2771-392X},
year = {2023},
date = {2023-01-01},
journal = {Applied Computing and Intelligence},
volume = {3},
number = {1},
pages = {44-78},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Piccoli, Michele; Zoni, Davide; Fornaciari, William; Massari, Giuseppe; Cococcioni, Marco; Rossi, Federico; Saponara, Sergio; Ruffaldi, Emanuele
Dynamic Power Consumption of the Full Posit Processing Unit: Analysis and Experiments Proceedings Article
In: Bispo, João; Charles, Henri-Pierre; Cherubin, Stefano; Massari, Giuseppe (Ed.): 14th Workshop on Parallel Programming and Run-Time Management Techniques for Many-Core Architectures and 12th Workshop on Design Tools and Architectures for Multicore Embedded Computing Platforms (PARMA-DITAM 2023), pp. 6:1–6:11, Schloss Dagstuhl – Leibniz-Zentrum für Informatik, Dagstuhl, Germany, 2023, ISSN: 2190-6807.
@inproceedings{piccoli_et_al:OASIcs.PARMA-DITAM.2023.6,
title = {Dynamic Power Consumption of the Full Posit Processing Unit: Analysis and Experiments},
author = {Michele Piccoli and Davide Zoni and William Fornaciari and Giuseppe Massari and Marco Cococcioni and Federico Rossi and Sergio Saponara and Emanuele Ruffaldi},
editor = {João Bispo and Henri-Pierre Charles and Stefano Cherubin and Giuseppe Massari},
url = {https://drops.dagstuhl.de/opus/volltexte/2023/17726},
doi = {10.4230/OASIcs.PARMA-DITAM.2023.6},
issn = {2190-6807},
year = {2023},
date = {2023-01-01},
booktitle = {14th Workshop on Parallel Programming and Run-Time Management Techniques for Many-Core Architectures and 12th Workshop on Design Tools and Architectures for Multicore Embedded Computing Platforms (PARMA-DITAM 2023)},
volume = {107},
pages = {6:1–6:11},
publisher = {Schloss Dagstuhl – Leibniz-Zentrum für Informatik},
address = {Dagstuhl, Germany},
series = {Open Access Series in Informatics (OASIcs)},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Nardini, Giovanni; Noferi, Alessandro; Ducange, Pietro; Stea, Giovanni
Exploiting Simu5G for generating datasets for training and testing AI models for 5G/6G network applications Journal Article
In: SoftwareX, vol. 21, pp. 101320, 2023, ISSN: 2352-7110.
@article{NARDINI2023101320,
title = {Exploiting Simu5G for generating datasets for training and testing AI models for 5G/6G network applications},
author = {Giovanni Nardini and Alessandro Noferi and Pietro Ducange and Giovanni Stea},
url = {https://www.sciencedirect.com/science/article/pii/S235271102300016X},
doi = {https://doi.org/10.1016/j.softx.2023.101320},
issn = {2352-7110},
year = {2023},
date = {2023-01-01},
journal = {SoftwareX},
volume = {21},
pages = {101320},
abstract = {Researchers working on Artificial Intelligence (AI) need suitable datasets for training and testing their models. When it comes to applications running through a mobile network, these datasets are difficult to obtain, because network operators are hardly willing to expose their network data or to open their network to experimentation. In this paper we show how Simu5G, a popular 5G network simulator based on OMNeT++, can be used to circumvent this problem: it allows users to log data at arbitrary spatial and temporal resolution, belonging to every network layer — from the application to the physical one.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Researchers working on Artificial Intelligence (AI) need suitable datasets for training and testing their models. When it comes to applications running through a mobile network, these datasets are difficult to obtain, because network operators are hardly willing to expose their network data or to open their network to experimentation. In this paper we show how Simu5G, a popular 5G network simulator based on OMNeT++, can be used to circumvent this problem: it allows users to log data at arbitrary spatial and temporal resolution, belonging to every network layer — from the application to the physical one.
Gustafson, John L.; Cococcioni, Marco; Rossi, Federico; Ruffaldi, Emanuele; Saponara, Sergio
Decoding-Free Two-Input Arithmetic for Low-Precision Real Numbers Proceedings Article
In: Gustafson, John; Leong, Siew Hoon; Michalewicz, Marek (Ed.): Next Generation Arithmetic, pp. 61–76, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-32180-1.
@inproceedings{10.1007/978-3-031-32180-1_4,
title = {Decoding-Free Two-Input Arithmetic for Low-Precision Real Numbers},
author = {John L. Gustafson and Marco Cococcioni and Federico Rossi and Emanuele Ruffaldi and Sergio Saponara},
editor = {John Gustafson and Siew Hoon Leong and Marek Michalewicz},
url = {https://dl.acm.org/doi/abs/10.1007/978-3-031-32180-1_4},
isbn = {978-3-031-32180-1},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {Next Generation Arithmetic},
pages = {61–76},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {In this work, we present a novel method for directly computing functions of two real numbers using logic circuits without decoding; the real numbers are mapped to a particularly-chosen set of integer numbers. We theoretically prove that this mapping always exists and that we can implement any kind of binary operation between real numbers regardless of the encoding format. While the real numbers in the set can be arbitrary (rational, irrational, transcendental), we find practical applications to low-precision posit™ number arithmetic. We finally provide examples for decoding-free 4-bit Posit arithmetic operations, showing a reduction in gate count up to a factor of $$7.6backslashtimes $$7.6×(and never below $$4.4backslashtimes $$4.4×) compared to a standard two-dimensional tabulation.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this work, we present a novel method for directly computing functions of two real numbers using logic circuits without decoding; the real numbers are mapped to a particularly-chosen set of integer numbers. We theoretically prove that this mapping always exists and that we can implement any kind of binary operation between real numbers regardless of the encoding format. While the real numbers in the set can be arbitrary (rational, irrational, transcendental), we find practical applications to low-precision posit™ number arithmetic. We finally provide examples for decoding-free 4-bit Posit arithmetic operations, showing a reduction in gate count up to a factor of $$7.6backslashtimes $$7.6×(and never below $$4.4backslashtimes $$4.4×) compared to a standard two-dimensional tabulation.
Paolini, Emilio; Marinis, Lorenzo De; Maggiani, Luca; Cococcioni, Marco; Andriolli, Nicola
CHARLES: A C++ fixed-point library for Photonic-Aware Neural Networks Journal Article
In: Neural Networks, vol. 162, pp. 531-540, 2023, ISSN: 0893-6080.
@article{PAOLINI2023531,
title = {CHARLES: A C++ fixed-point library for Photonic-Aware Neural Networks},
author = {Emilio Paolini and Lorenzo De Marinis and Luca Maggiani and Marco Cococcioni and Nicola Andriolli},
url = {https://www.sciencedirect.com/science/article/pii/S0893608023001247},
doi = {https://doi.org/10.1016/j.neunet.2023.03.007},
issn = {0893-6080},
year = {2023},
date = {2023-01-01},
journal = {Neural Networks},
volume = {162},
pages = {531-540},
abstract = {In this paper we present CHARLES (C++ pHotonic Aware neuRaL nEtworkS), a C++ library aimed at providing a flexible tool to simulate the behavior of Photonic-Aware Neural Network (PANN). PANNs are neural network architectures aware of the constraints due to the underlying photonic hardware, mostly in terms of low equivalent precision of the computations. For this reason, CHARLES exploits fixed-point computations for inference, while it supports both floating-point and fixed-point numerical formats for training. In this way, we can compare the effects due to the quantization in the inference phase when the training phase is performed on a classical floating-point model and on a model exploiting high-precision fixed-point numbers. To validate CHARLES and identify the most suited numerical format for PANN training, we report the simulation results obtained considering three datasets: Iris, MNIST, and Fashion-MNIST. Fixed-training is shown to outperform floating-training when executing inference on bitwidths suitable for photonic implementation. Indeed, performing the training phase in the floating-point domain and then quantizing to lower bitwidths results in a very high accuracy loss. Instead, when fixed-point numbers are exploited in the training phase, the accuracy loss due to quantization to lower bitwidths is significantly reduced. In particular, we show that for Iris dataset, fixed-training achieves a performance similar to floating-training. Fixed-training allows to obtain an accuracy of 90.4% and 68.1% with the MNIST and Fashion-MNIST datasets using only 6 bits, while the floating-training reaches an accuracy of just 25.4% and 50.0% when exploiting the same bitwidths.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper we present CHARLES (C++ pHotonic Aware neuRaL nEtworkS), a C++ library aimed at providing a flexible tool to simulate the behavior of Photonic-Aware Neural Network (PANN). PANNs are neural network architectures aware of the constraints due to the underlying photonic hardware, mostly in terms of low equivalent precision of the computations. For this reason, CHARLES exploits fixed-point computations for inference, while it supports both floating-point and fixed-point numerical formats for training. In this way, we can compare the effects due to the quantization in the inference phase when the training phase is performed on a classical floating-point model and on a model exploiting high-precision fixed-point numbers. To validate CHARLES and identify the most suited numerical format for PANN training, we report the simulation results obtained considering three datasets: Iris, MNIST, and Fashion-MNIST. Fixed-training is shown to outperform floating-training when executing inference on bitwidths suitable for photonic implementation. Indeed, performing the training phase in the floating-point domain and then quantizing to lower bitwidths results in a very high accuracy loss. Instead, when fixed-point numbers are exploited in the training phase, the accuracy loss due to quantization to lower bitwidths is significantly reduced. In particular, we show that for Iris dataset, fixed-training achieves a performance similar to floating-training. Fixed-training allows to obtain an accuracy of 90.4% and 68.1% with the MNIST and Fashion-MNIST datasets using only 6 bits, while the floating-training reaches an accuracy of just 25.4% and 50.0% when exploiting the same bitwidths.
Dell'Oglio, Pietro; Bondielli, Alessandro; Bechini, Alessio; Marcelloni, Francesco
Leveraging Sequence Mining for Robot Process Automation Proceedings Article
In: Abraham, Ajith; Pllana, Sabri; Casalino, Gabriella; Ma, Kun; Bajaj, Anu (Ed.): Intelligent Systems Design and Applications, pp. 224–233, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-35510-3.
@inproceedings{10.1007/978-3-031-35510-3_22,
title = {Leveraging Sequence Mining for Robot Process Automation},
author = {Pietro Dell'Oglio and Alessandro Bondielli and Alessio Bechini and Francesco Marcelloni},
editor = {Ajith Abraham and Sabri Pllana and Gabriella Casalino and Kun Ma and Anu Bajaj},
isbn = {978-3-031-35510-3},
year = {2023},
date = {2023-01-01},
booktitle = {Intelligent Systems Design and Applications},
pages = {224–233},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {The automation of sequences of repetitive actions performed by human operators in interacting with software applications is crucial to prevent work from being perceived as alienating and boring. Robot applications can automatise these sequences once they have been identified. In this paper, we propose a two-step approach to mine sequences of actions that could be automated from log data produced by the interactions of a human operator with specific software applications. Since the number of possible sequences may be very high and not all the sequences are interesting to be automatised, we focus our mining process on sequences that meet precise patterns. First, Frequent Episode Mining algorithms are applied for extracting all the sequences of actions that occur with at least a minimum frequency. Then, we exploit fuzzy string matching based on the Levenshtein distance for filtering out the sequences that do not match established patterns. We evaluate the effectiveness of the approach using a benchmark dataset and present a case study on a real-world dataset of activity logs generated in the context of the AUTOMIA project.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The automation of sequences of repetitive actions performed by human operators in interacting with software applications is crucial to prevent work from being perceived as alienating and boring. Robot applications can automatise these sequences once they have been identified. In this paper, we propose a two-step approach to mine sequences of actions that could be automated from log data produced by the interactions of a human operator with specific software applications. Since the number of possible sequences may be very high and not all the sequences are interesting to be automatised, we focus our mining process on sequences that meet precise patterns. First, Frequent Episode Mining algorithms are applied for extracting all the sequences of actions that occur with at least a minimum frequency. Then, we exploit fuzzy string matching based on the Levenshtein distance for filtering out the sequences that do not match established patterns. We evaluate the effectiveness of the approach using a benchmark dataset and present a case study on a real-world dataset of activity logs generated in the context of the AUTOMIA project.
Ferrante, Nicola; Giuffrida, Gianluca; Nannipieri, Pietro; Bechini, Alessio; Fanucci, Luca
Fault Detection Exploiting Artificial Intelligence in Satellite Systems Proceedings Article
In: Ieracitano, Cosimo; Mammone, Nadia; Clemente, Marco Di; Mahmud, Mufti; Furfaro, Roberto; Morabito, Francesco Carlo (Ed.): The Use of Artificial Intelligence for Space Applications, pp. 151–166, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-25755-1.
@inproceedings{10.1007/978-3-031-25755-1_10,
title = {Fault Detection Exploiting Artificial Intelligence in Satellite Systems},
author = {Nicola Ferrante and Gianluca Giuffrida and Pietro Nannipieri and Alessio Bechini and Luca Fanucci},
editor = {Cosimo Ieracitano and Nadia Mammone and Marco Di Clemente and Mufti Mahmud and Roberto Furfaro and Francesco Carlo Morabito},
isbn = {978-3-031-25755-1},
year = {2023},
date = {2023-01-01},
booktitle = {The Use of Artificial Intelligence for Space Applications},
pages = {151–166},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Mission control and fault management are fundamental in safety-critical scenarios such as space applications. To this extent, fault detection techniques are crucial to meet the desired safety and integrity level. This work proposes a fault detection system exploiting an autoregressive model, which is based on a Deep Neural Network (DNN). We trained the aforementioned model on a dataset composed of telemetries acquired from Mars Advanced Radar for Subsurface and Ionosphere Sounding (MARSIS). The training process has been designed as a sequence-to-sequence task, varying the length of input and output time series. Several DNN architectures were proposed, using both Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) as basic building blocks. Lastly, we performed fault injection modeling faults of different nature. The results obtained show that the proposed solution detects up to 90% of injected faults. We found that GRU-based models outperform LSTM-based ones in this task. Furthermore, we demonstrated that we can predict signal evolution without any knowledge of the underlying physics of the system, substituting a DNN to the traditional differential equations, reducing expertise and time-to-market concerning existing solutions.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Mission control and fault management are fundamental in safety-critical scenarios such as space applications. To this extent, fault detection techniques are crucial to meet the desired safety and integrity level. This work proposes a fault detection system exploiting an autoregressive model, which is based on a Deep Neural Network (DNN). We trained the aforementioned model on a dataset composed of telemetries acquired from Mars Advanced Radar for Subsurface and Ionosphere Sounding (MARSIS). The training process has been designed as a sequence-to-sequence task, varying the length of input and output time series. Several DNN architectures were proposed, using both Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) as basic building blocks. Lastly, we performed fault injection modeling faults of different nature. The results obtained show that the proposed solution detects up to 90% of injected faults. We found that GRU-based models outperform LSTM-based ones in this task. Furthermore, we demonstrated that we can predict signal evolution without any knowledge of the underlying physics of the system, substituting a DNN to the traditional differential equations, reducing expertise and time-to-market concerning existing solutions.
Sansone, Giacomo; Cococcioni, Marco
Experiments on Speeding Up the Recursive Fast Fourier Transform by Using AVX-512 SIMD Instructions Proceedings Article
In: Berta, Riccardo; Gloria, Alessandro De (Ed.): Applications in Electronics Pervading Industry, Environment and Society, pp. 255–263, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-30333-3.
@inproceedings{10.1007/978-3-031-30333-3_34,
title = {Experiments on Speeding Up the Recursive Fast Fourier Transform by Using AVX-512 SIMD Instructions},
author = {Giacomo Sansone and Marco Cococcioni},
editor = {Riccardo Berta and Alessandro De Gloria},
isbn = {978-3-031-30333-3},
year = {2023},
date = {2023-01-01},
booktitle = {Applications in Electronics Pervading Industry, Environment and Society},
pages = {255–263},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {The Fast Fourier Transform is probably one of the most studied algorithms of all time. New techniques regarding hardware and software are often applied and tested on it, but the interest in FFT is still large because of its applications - signal and image processing, numerical computations, etc. In this paper, we start from a trivial recursive version of the algorithm and we speed it up using AVX-512 Single Instruction Multiple Data (SIMD) instructions on an Intel i7 CPU with native support to AVX-512. In particular, we study the impact of two different storage choices of vector of complex numbers: block interleaving and complex interleaving. Experimental results show that automatic vectorization provides a 10.65% ($$backslashsim 1.12backslashtimes $$∼1.12×) speedup, while with vectorization done by hand the speedup reaches 33.78% ($$backslashsim 1.51backslashtimes $$∼1.51×). We have made our code publicly available, which could be helpful for SIMD instructions teaching purposes.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The Fast Fourier Transform is probably one of the most studied algorithms of all time. New techniques regarding hardware and software are often applied and tested on it, but the interest in FFT is still large because of its applications - signal and image processing, numerical computations, etc. In this paper, we start from a trivial recursive version of the algorithm and we speed it up using AVX-512 Single Instruction Multiple Data (SIMD) instructions on an Intel i7 CPU with native support to AVX-512. In particular, we study the impact of two different storage choices of vector of complex numbers: block interleaving and complex interleaving. Experimental results show that automatic vectorization provides a 10.65% ($$backslashsim 1.12backslashtimes $$∼1.12×) speedup, while with vectorization done by hand the speedup reaches 33.78% ($$backslashsim 1.51backslashtimes $$∼1.51×). We have made our code publicly available, which could be helpful for SIMD instructions teaching purposes.
Donati, Massimiliano; Bechini, Alessio; D'Anna, Clelia; Fattori, Bruno; Marini, Marco; Olivelli, Martina; Pelagatti, Susanna; Ricci, Giulia; Schirinzi, Erika; Siciliano, Gabriele; Tavosanis, Mirko; Torri, Francesca; Vanello, Nicola; Fanucci, Luca
A Clinical Tool for Prognosis and Speech Rehabilitation in Dysarthric Patients: The DESIRE Project Proceedings Article
In: Berta, Riccardo; Gloria, Alessandro De (Ed.): Applications in Electronics Pervading Industry, Environment and Society, pp. 380–385, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-30333-3.
@inproceedings{10.1007/978-3-031-30333-3_52,
title = {A Clinical Tool for Prognosis and Speech Rehabilitation in Dysarthric Patients: The DESIRE Project},
author = {Massimiliano Donati and Alessio Bechini and Clelia D'Anna and Bruno Fattori and Marco Marini and Martina Olivelli and Susanna Pelagatti and Giulia Ricci and Erika Schirinzi and Gabriele Siciliano and Mirko Tavosanis and Francesca Torri and Nicola Vanello and Luca Fanucci},
editor = {Riccardo Berta and Alessandro De Gloria},
isbn = {978-3-031-30333-3},
year = {2023},
date = {2023-01-01},
booktitle = {Applications in Electronics Pervading Industry, Environment and Society},
pages = {380–385},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Dysarthria is a motor disorder of speech characterized by alteration of articulation and intelligibility of speech. The goal of dysarthria management is to optimize communication effectiveness for as long as possible. To help clinicians in monitoring disease progression and rehabilitation outcomes, the DESIRE tool analyzes several reading sessions in which the patients pronounce predetermined selected words aloud, elaborating a measure of how much the patient's pronunciation deviates from those of previous sessions and the expected performance. In addition, the electronical record offers a comprehensive view of patient's status, and the web access allows the care team to remotely monitor progresses, so that they can tailor rehabilitation programs over time. Through the possibility to understand the patient difficulty about specific phonemes, word length, consonant clusters, this innovative tool offers a method to assess and monitoring dysarthria, to address therapeutic strategies, and to provide useful requirements for clinical trials readiness.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Dysarthria is a motor disorder of speech characterized by alteration of articulation and intelligibility of speech. The goal of dysarthria management is to optimize communication effectiveness for as long as possible. To help clinicians in monitoring disease progression and rehabilitation outcomes, the DESIRE tool analyzes several reading sessions in which the patients pronounce predetermined selected words aloud, elaborating a measure of how much the patient's pronunciation deviates from those of previous sessions and the expected performance. In addition, the electronical record offers a comprehensive view of patient's status, and the web access allows the care team to remotely monitor progresses, so that they can tailor rehabilitation programs over time. Through the possibility to understand the patient difficulty about specific phonemes, word length, consonant clusters, this innovative tool offers a method to assess and monitoring dysarthria, to address therapeutic strategies, and to provide useful requirements for clinical trials readiness.
Rossi, Federico; Fiaschi, Lorenzo; Cococcioni, Marco; Saponara, Sergio
Design and FPGA Synthesis of BAN Processing Unit for Non-Archimedean Number Crunching Proceedings Article
In: Berta, Riccardo; Gloria, Alessandro De (Ed.): Applications in Electronics Pervading Industry, Environment and Society, pp. 320–325, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-30333-3.
@inproceedings{10.1007/978-3-031-30333-3_43,
title = {Design and FPGA Synthesis of BAN Processing Unit for Non-Archimedean Number Crunching},
author = {Federico Rossi and Lorenzo Fiaschi and Marco Cococcioni and Sergio Saponara},
editor = {Riccardo Berta and Alessandro De Gloria},
isbn = {978-3-031-30333-3},
year = {2023},
date = {2023-01-01},
booktitle = {Applications in Electronics Pervading Industry, Environment and Society},
pages = {320–325},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {This work presents the design and synthesis of a processing unit for numbers encoded according to the recently introduced BAN format. Such an encoding allows one to represent numbers which are not only finite (as the reals) but also infinitely large or infinitely small, i.e., non-Archimedean. The motivation behind this study is the significant burst the non-Archimedean numerical computations have received in the last 20 years and the applications that have been found. With a hardware support, this operations would significantly increase in speed, enlarging the spectrum of possible applications to industrial and real-time ones.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This work presents the design and synthesis of a processing unit for numbers encoded according to the recently introduced BAN format. Such an encoding allows one to represent numbers which are not only finite (as the reals) but also infinitely large or infinitely small, i.e., non-Archimedean. The motivation behind this study is the significant burst the non-Archimedean numerical computations have received in the last 20 years and the applications that have been found. With a hardware support, this operations would significantly increase in speed, enlarging the spectrum of possible applications to industrial and real-time ones.
Casalino, Gabriella; Ducange, Pietro; Fazzolari, Michela; Pecori, Riccardo
Incremental and Interpretable Learning Analytics Through Fuzzy Hoeffding Decision Trees Proceedings Article
In: Fulantelli, Giovanni; Burgos, Daniel; Casalino, Gabriella; Cimitile, Marta; Bosco, Giosuè Lo; Taibi, Davide (Ed.): Higher Education Learning Methodologies and Technologies Online, pp. 674–690, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-29800-4.
@inproceedings{10.1007/978-3-031-29800-4_51,
title = {Incremental and Interpretable Learning Analytics Through Fuzzy Hoeffding Decision Trees},
author = {Gabriella Casalino and Pietro Ducange and Michela Fazzolari and Riccardo Pecori},
editor = {Giovanni Fulantelli and Daniel Burgos and Gabriella Casalino and Marta Cimitile and Giosuè Lo Bosco and Davide Taibi},
isbn = {978-3-031-29800-4},
year = {2023},
date = {2023-01-01},
booktitle = {Higher Education Learning Methodologies and Technologies Online},
pages = {674–690},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Artificial Intelligence-based methods have been thoroughly applied in various fields over the years and the educational scenario is not an exception. However, the usage of the so-called explainable Artificial Intelligence, even if desirable, is still limited, especially whenever we consider educational datasets. Moreover, the time dimension is not often regarded enough when analyzing such types of data. In this paper, we have applied the fuzzy version of the Hoeffding Decision Tree to an educational dataset, considering separately STEM and Social Sciences subjects, in order to take into consideration both the time evolution of the educational process and the possible interpretability of the final results. The considered models resulted to be successful in discriminating the passing or failing of exams at the end of consecutive semesters on the part of students. Moreover, Fuzzy Hoeffding Decision Tree occurred to be much more compact and interpretable compared to the traditional Hoeffding Decision Tree.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Artificial Intelligence-based methods have been thoroughly applied in various fields over the years and the educational scenario is not an exception. However, the usage of the so-called explainable Artificial Intelligence, even if desirable, is still limited, especially whenever we consider educational datasets. Moreover, the time dimension is not often regarded enough when analyzing such types of data. In this paper, we have applied the fuzzy version of the Hoeffding Decision Tree to an educational dataset, considering separately STEM and Social Sciences subjects, in order to take into consideration both the time evolution of the educational process and the possible interpretability of the final results. The considered models resulted to be successful in discriminating the passing or failing of exams at the end of consecutive semesters on the part of students. Moreover, Fuzzy Hoeffding Decision Tree occurred to be much more compact and interpretable compared to the traditional Hoeffding Decision Tree.
Fioriti, Davide; Stevanato, Nicolò; Ducange, Pietro; Marcelloni, Francesco; Colombo, Emanuela; Poli, Davide
In: IEEE Access, pp. 1-1, 2023.
@article{10179910,
title = {Data platform guidelines and prototype for microgrids and energy access: matching demand profiles and socio-economic data to foster project development},
author = {Davide Fioriti and Nicolò Stevanato and Pietro Ducange and Francesco Marcelloni and Emanuela Colombo and Davide Poli},
doi = {10.1109/ACCESS.2023.3294841},
year = {2023},
date = {2023-01-01},
journal = {IEEE Access},
pages = {1-1},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Daole, Mattia; Schiavo, Alessio; Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro
OpenFL-XAI: Federated learning of explainable artificial intelligence models in Python Journal Article
In: SoftwareX, vol. 23, pp. 101505, 2023, ISSN: 2352-7110.
@article{DAOLE2023101505,
title = {OpenFL-XAI: Federated learning of explainable artificial intelligence models in Python},
author = {Mattia Daole and Alessio Schiavo and José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Alessandro Renda},
url = {https://www.sciencedirect.com/science/article/pii/S2352711023002017},
doi = {https://doi.org/10.1016/j.softx.2023.101505},
issn = {2352-7110},
year = {2023},
date = {2023-01-01},
journal = {SoftwareX},
volume = {23},
pages = {101505},
abstract = {Artificial Intelligence (AI) systems play a significant role in manifold decision-making processes in our daily lives, making trustworthiness of AI more and more crucial for its widespread acceptance. Among others, privacy and explainability are considered key requirements for enabling trust in AI. Building on these needs, we propose a software for Federated Learning (FL) of Rule-Based Systems (RBSs): on one hand FL prioritizes user data privacy during collaborative model training. On the other hand, RBSs are deemed as interpretable-by-design models and ensure high transparency in the decision-making process. The proposed software, developed as an extension to the Intel® OpenFL open-source framework, offers a viable solution for developing AI applications balancing accuracy, privacy, and interpretability.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Artificial Intelligence (AI) systems play a significant role in manifold decision-making processes in our daily lives, making trustworthiness of AI more and more crucial for its widespread acceptance. Among others, privacy and explainability are considered key requirements for enabling trust in AI. Building on these needs, we propose a software for Federated Learning (FL) of Rule-Based Systems (RBSs): on one hand FL prioritizes user data privacy during collaborative model training. On the other hand, RBSs are deemed as interpretable-by-design models and ensure high transparency in the decision-making process. The proposed software, developed as an extension to the Intel® OpenFL open-source framework, offers a viable solution for developing AI applications balancing accuracy, privacy, and interpretability.
Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Nardini, Giovanni; Noferi, Alessandro; Renda, Alessandro; Ruffini, Fabrizio; Schiavo, Alessio; Stea, Giovanni; Virdis, Antonio
Enabling federated learning of explainable AI models within beyond-5G/6G networks Journal Article
In: Computer Communications, 2023, ISSN: 0140-3664.
@article{BARCENA2023,
title = {Enabling federated learning of explainable AI models within beyond-5G/6G networks},
author = {José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Giovanni Nardini and Alessandro Noferi and Alessandro Renda and Fabrizio Ruffini and Alessio Schiavo and Giovanni Stea and Antonio Virdis},
url = {https://www.sciencedirect.com/science/article/pii/S0140366423002724},
doi = {https://doi.org/10.1016/j.comcom.2023.07.039},
issn = {0140-3664},
year = {2023},
date = {2023-01-01},
journal = {Computer Communications},
abstract = {The quest for trustworthiness in Artificial Intelligence (AI) is increasingly urgent, especially in the field of next-generation wireless networks. Future Beyond 5G (B5G)/6G networks will connect a huge amount of devices and will offer innovative services empowered with AI and Machine Learning tools. Nevertheless, private user data, which are essential for training such services, are not an asset that can be unrestrictedly shared over the network, mainly because of privacy concerns. To overcome this issue, Federated Learning (FL) has recently been proposed as a paradigm to enable collaborative model training among multiple parties, without any disclosure of private raw data. However, the initiative to natively integrate FL services into mobile networks is still far from being accomplished. In this paper we propose a novel FL-as-a-Service framework that provides the B5G/6G network with flexible mechanisms to allow end users to exploit FL services, and we describe its applicability to a Quality of Experience (QoE) forecasting service based on a vehicular networking use case. Specifically, we show how FL of eXplainable AI (XAI) models can be leveraged for the QoE forecasting task, and induces a benefit in terms of both accuracy, compared to local learning, and trustworthiness, thanks to the adoption of inherently interpretable models. Such considerations are supported by an extensive experimental analysis on a publicly available simulated dataset. Finally, we assessed how the learning process is affected by the system deployment and the performance of the underlying communication and computation infrastructure, through system-level simulations, which show the benefits of deploying the proposed framework in edge-based environments.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The quest for trustworthiness in Artificial Intelligence (AI) is increasingly urgent, especially in the field of next-generation wireless networks. Future Beyond 5G (B5G)/6G networks will connect a huge amount of devices and will offer innovative services empowered with AI and Machine Learning tools. Nevertheless, private user data, which are essential for training such services, are not an asset that can be unrestrictedly shared over the network, mainly because of privacy concerns. To overcome this issue, Federated Learning (FL) has recently been proposed as a paradigm to enable collaborative model training among multiple parties, without any disclosure of private raw data. However, the initiative to natively integrate FL services into mobile networks is still far from being accomplished. In this paper we propose a novel FL-as-a-Service framework that provides the B5G/6G network with flexible mechanisms to allow end users to exploit FL services, and we describe its applicability to a Quality of Experience (QoE) forecasting service based on a vehicular networking use case. Specifically, we show how FL of eXplainable AI (XAI) models can be leveraged for the QoE forecasting task, and induces a benefit in terms of both accuracy, compared to local learning, and trustworthiness, thanks to the adoption of inherently interpretable models. Such considerations are supported by an extensive experimental analysis on a publicly available simulated dataset. Finally, we assessed how the learning process is affected by the system deployment and the performance of the underlying communication and computation infrastructure, through system-level simulations, which show the benefits of deploying the proposed framework in edge-based environments.
Bondielli, A.; Dell'Oglio, P.; Lenci, A.; Marcelloni, F.; Passaro, L. C.; Sabbatini, M.
EVALITA 2023: 8th Evaluation Campaign of Natural Language Processing and Speech Tools for Italian, Sep 7-8, Parma, IT, vol. 3473, 2023.
@conference{bondielli2023,
title = {Multi-fake-detective at evalita 2023: Overview of the multimodal fake news detection and verification task},
author = {A. Bondielli and P. Dell'Oglio and A. Lenci and F. Marcelloni and L. C. Passaro and M. Sabbatini},
url = {https://ceur-ws.org/Vol-3473/paper32.pdf},
year = {2023},
date = {2023-01-01},
urldate = {2023-01-01},
booktitle = {EVALITA 2023: 8th Evaluation Campaign of Natural Language Processing and Speech Tools for Italian, Sep 7-8, Parma, IT},
journal = {CEUR Workshop Proceedings},
volume = {3473},
abstract = {This paper introduces the MULTI-Fake-DetectiVE shared task for the EVALITA 2023 campaign. The task was aimed at exploring multimodality within the realm of fake news and intended to address the problem from two perspectives, represented by the two sub-tasks. In sub-task 1, we aimed to evaluate the effectiveness of multimodal fake news detection systems. In sub-task 2, we sought to gain insights into the interplay between text and images, specifically how they mutually influence the interpretation of content in the context of distinguishing between fake and real news. Both perspectives were framed as classification problems. The paper presents an overview of the task. In particular, we detail the key aspects of the task, including the creation of a new dataset for fake news detection in Italian, the evaluation methodology and criteria, the participant systems, and their results. In light of the obtained results, we argue that the problem is still open and propose some future directions.},
keywords = {},
pubstate = {published},
tppubtype = {conference}
}
This paper introduces the MULTI-Fake-DetectiVE shared task for the EVALITA 2023 campaign. The task was aimed at exploring multimodality within the realm of fake news and intended to address the problem from two perspectives, represented by the two sub-tasks. In sub-task 1, we aimed to evaluate the effectiveness of multimodal fake news detection systems. In sub-task 2, we sought to gain insights into the interplay between text and images, specifically how they mutually influence the interpretation of content in the context of distinguishing between fake and real news. Both perspectives were framed as classification problems. The paper presents an overview of the task. In particular, we detail the key aspects of the task, including the creation of a new dataset for fake news detection in Italian, the evaluation methodology and criteria, the participant systems, and their results. In light of the obtained results, we argue that the problem is still open and propose some future directions.
Dell’Oglio, Pietro; Bondielli, Alessandro; Marcelloni, Francesco
A System to Support Readers in Automatically Acquiring Complete Summarized Information on an Event from Different Sources Journal Article
In: Algorithms, vol. 16, no. 11, 2023, ISSN: 1999-4893.
@article{a16110513,
title = {A System to Support Readers in Automatically Acquiring Complete Summarized Information on an Event from Different Sources},
author = {Pietro Dell’Oglio and Alessandro Bondielli and Francesco Marcelloni},
url = {https://www.mdpi.com/1999-4893/16/11/513},
doi = {10.3390/a16110513},
issn = {1999-4893},
year = {2023},
date = {2023-01-01},
journal = {Algorithms},
volume = {16},
number = {11},
abstract = {Today, most newspapers utilize social media to disseminate news. On the one hand, this results in an overload of related articles for social media users. On the other hand, since social media tends to form echo chambers around their users, different opinions and information may be hidden. Enabling users to access different information (possibly outside of their echo chambers, without the burden of reading entire articles, often containing redundant information) may be a step forward in allowing them to form their own opinions. To address this challenge, we propose a system that integrates Transformer neural models and text summarization models along with decision rules. Given a reference article already read by the user, our system first collects articles related to the same topic from a configurable number of different sources. Then, it identifies and summarizes the information that differs from the reference article and outputs the summary to the user. The core of the system is the sentence classification algorithm, which classifies sentences in the collected articles into three classes based on similarity with the reference article: sentences classified as dissimilar are summarized by using a pre-trained abstractive summarization model. We evaluated the proposed system in two steps. First, we assessed its effectiveness in identifying content differences between the reference article and the related articles by using human judgments obtained through crowdsourcing as ground truth. We obtained an average F1 score of 0.772 against average F1 scores of 0.797 and 0.676 achieved by two state-of-the-art approaches based, respectively, on model tuning and prompt tuning, which require an appropriate tuning phase and, therefore, greater computational effort. Second, we asked a sample of people to evaluate how well the summary generated by the system represents the information that is not present in the article read by the user. The results are extremely encouraging. Finally, we present a use case.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Today, most newspapers utilize social media to disseminate news. On the one hand, this results in an overload of related articles for social media users. On the other hand, since social media tends to form echo chambers around their users, different opinions and information may be hidden. Enabling users to access different information (possibly outside of their echo chambers, without the burden of reading entire articles, often containing redundant information) may be a step forward in allowing them to form their own opinions. To address this challenge, we propose a system that integrates Transformer neural models and text summarization models along with decision rules. Given a reference article already read by the user, our system first collects articles related to the same topic from a configurable number of different sources. Then, it identifies and summarizes the information that differs from the reference article and outputs the summary to the user. The core of the system is the sentence classification algorithm, which classifies sentences in the collected articles into three classes based on similarity with the reference article: sentences classified as dissimilar are summarized by using a pre-trained abstractive summarization model. We evaluated the proposed system in two steps. First, we assessed its effectiveness in identifying content differences between the reference article and the related articles by using human judgments obtained through crowdsourcing as ground truth. We obtained an average F1 score of 0.772 against average F1 scores of 0.797 and 0.676 achieved by two state-of-the-art approaches based, respectively, on model tuning and prompt tuning, which require an appropriate tuning phase and, therefore, greater computational effort. Second, we asked a sample of people to evaluate how well the summary generated by the system represents the information that is not present in the article read by the user. The results are extremely encouraging. Finally, we present a use case.
Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro; Ruffini, Fabrizio; Schiavo, Alessio
Federated TSK Models for Predicting Quality of Experience in B5G/6G Networks Proceedings Article
In: 2023 IEEE International Conference on Fuzzy Systems (FUZZ), pp. 1-8, 2023.
@inproceedings{10309758,
title = {Federated TSK Models for Predicting Quality of Experience in B5G/6G Networks},
author = {José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Alessandro Renda and Fabrizio Ruffini and Alessio Schiavo},
doi = {10.1109/FUZZ52849.2023.10309758},
year = {2023},
date = {2023-01-01},
booktitle = {2023 IEEE International Conference on Fuzzy Systems (FUZZ)},
pages = {1-8},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Bechini, Alessio; Daole, Mattia; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro
An Application for Federated Learning of XAI Models in Edge Computing Environments Proceedings Article
In: 2023 IEEE International Conference on Fuzzy Systems (FUZZ), pp. 1-7, 2023.
@inproceedings{10309783,
title = {An Application for Federated Learning of XAI Models in Edge Computing Environments},
author = {Alessio Bechini and Mattia Daole and Pietro Ducange and Francesco Marcelloni and Alessandro Renda},
doi = {10.1109/FUZZ52849.2023.10309783},
year = {2023},
date = {2023-01-01},
booktitle = {2023 IEEE International Conference on Fuzzy Systems (FUZZ)},
pages = {1-7},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro; Ruffini, Fabrizio
Federated Learning of Explainable Artificial Intelligence Models for Predicting Parkinson's Disease Progression Proceedings Article
In: Longo, Luca (Ed.): Explainable Artificial Intelligence, pp. 630–648, Springer Nature Switzerland, Cham, 2023, ISBN: 978-3-031-44064-9.
@inproceedings{10.1007/978-3-031-44064-9_34,
title = {Federated Learning of Explainable Artificial Intelligence Models for Predicting Parkinson's Disease Progression},
author = {José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Alessandro Renda and Fabrizio Ruffini},
editor = {Luca Longo},
isbn = {978-3-031-44064-9},
year = {2023},
date = {2023-01-01},
booktitle = {Explainable Artificial Intelligence},
pages = {630–648},
publisher = {Springer Nature Switzerland},
address = {Cham},
abstract = {Services based on Artificial Intelligence (AI) are becoming increasingly pervasive in our society. At the same time, however, we are also witnessing a growing awareness towards the ethical aspects and the trustworthiness of AI tools, especially in high stakes domains, such as the healthcare one. In this paper, we propose the adoption of AI techniques for predicting Parkinson's Disease progression with the overarching aim of accommodating the urgent need for trustworthiness. We address two key requirements towards trustworthy AI, namely privacy preservation in learning AI models and their explainability. As for the former aspect, we consider the (rather common) case of medical data coming from different health institutions, assuming that they cannot be shared due to privacy concerns. To address this shortcoming, we leverage federated learning (FL) as a paradigm for collaborative model training among multiple parties without any disclosure of private raw data. As for the latter aspect, we focus on highly interpretable models, i.e., those for which humans are able to understand how decisions have been taken. An extensive experimental analysis carried out on a well-known Parkinson Telemonitoring dataset highlights how the proposed approach based on FL of fuzzy rule-based systems allows achieving, simultaneously, data privacy and interpretability. Results are reported for different data partitioning scenarios, also comparing the interpretable-by-design model with an opaque neural network model.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Services based on Artificial Intelligence (AI) are becoming increasingly pervasive in our society. At the same time, however, we are also witnessing a growing awareness towards the ethical aspects and the trustworthiness of AI tools, especially in high stakes domains, such as the healthcare one. In this paper, we propose the adoption of AI techniques for predicting Parkinson's Disease progression with the overarching aim of accommodating the urgent need for trustworthiness. We address two key requirements towards trustworthy AI, namely privacy preservation in learning AI models and their explainability. As for the former aspect, we consider the (rather common) case of medical data coming from different health institutions, assuming that they cannot be shared due to privacy concerns. To address this shortcoming, we leverage federated learning (FL) as a paradigm for collaborative model training among multiple parties without any disclosure of private raw data. As for the latter aspect, we focus on highly interpretable models, i.e., those for which humans are able to understand how decisions have been taken. An extensive experimental analysis carried out on a well-known Parkinson Telemonitoring dataset highlights how the proposed approach based on FL of fuzzy rule-based systems allows achieving, simultaneously, data privacy and interpretability. Results are reported for different data partitioning scenarios, also comparing the interpretable-by-design model with an opaque neural network model.
2022
Bechini, Alessio; Bondielli, Alessandro; Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro
A News-Based Framework for Uncovering and Tracking City Area Profiles: Assessment in Covid-19 Setting Journal Article
In: ACM Trans. Knowl. Discov. Data, vol. 16, no. 6, 2022, ISSN: 1556-4681.
@article{10.1145/3532186,
title = {A News-Based Framework for Uncovering and Tracking City Area Profiles: Assessment in Covid-19 Setting},
author = {Alessio Bechini and Alessandro Bondielli and José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Alessandro Renda},
url = {https://doi.org/10.1145/3532186},
doi = {10.1145/3532186},
issn = {1556-4681},
year = {2022},
date = {2022-07-01},
urldate = {2022-07-01},
journal = {ACM Trans. Knowl. Discov. Data},
volume = {16},
number = {6},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
abstract = {In the last years, there has been an ever-increasing interest in profiling various aspects of city life, especially in the context of smart cities. This interest has become even more relevant recently when we have realized how dramatic events, such as the Covid-19 pandemic, can deeply affect the city life, producing drastic changes. Identifying and analyzing such changes, both at the city level and within single neighborhoods, may be a fundamental tool to better manage the current situation and provide sound strategies for future planning. Furthermore, such fine-grained and up-to-date characterization can represent a valuable asset for other tools and services, e.g., web mapping applications or real estate agency platforms. In this article, we propose a framework featuring a novel methodology to model and track changes in areas of the city by extracting information from online newspaper articles. The problem of uncovering clusters of news at specific times is tackled by means of the joint use of state-of-the-art language models to represent the articles, and of a density-based streaming clustering algorithm, properly shaped to deal with high-dimensional text embeddings. Furthermore, we propose a method to automatically label the obtained clusters in a semantically meaningful way, and we introduce a set of metrics aimed at tracking the temporal evolution of clusters. A case study focusing on the city of Rome during the Covid-19 pandemic is illustrated and discussed to evaluate the effectiveness of the proposed approach.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In the last years, there has been an ever-increasing interest in profiling various aspects of city life, especially in the context of smart cities. This interest has become even more relevant recently when we have realized how dramatic events, such as the Covid-19 pandemic, can deeply affect the city life, producing drastic changes. Identifying and analyzing such changes, both at the city level and within single neighborhoods, may be a fundamental tool to better manage the current situation and provide sound strategies for future planning. Furthermore, such fine-grained and up-to-date characterization can represent a valuable asset for other tools and services, e.g., web mapping applications or real estate agency platforms. In this article, we propose a framework featuring a novel methodology to model and track changes in areas of the city by extracting information from online newspaper articles. The problem of uncovering clusters of news at specific times is tackled by means of the joint use of state-of-the-art language models to represent the articles, and of a density-based streaming clustering algorithm, properly shaped to deal with high-dimensional text embeddings. Furthermore, we propose a method to automatically label the obtained clusters in a semantically meaningful way, and we introduce a set of metrics aimed at tracking the temporal evolution of clusters. A case study focusing on the city of Rome during the Covid-19 pandemic is illustrated and discussed to evaluate the effectiveness of the proposed approach.
Renda, Alessandro; Ducange, Pietro; Marcelloni, Francesco; Sabella, Dario; Filippou, Miltiadis C.; Nardini, Giovanni; Stea, Giovanni; Virdis, Antonio; Micheli, Davide; Rapone, Damiano; Baltar, Leonardo Gomes
Federated Learning of Explainable AI Models in 6G Systems: Towards Secure and Automated Vehicle Networking Journal Article
In: Information, vol. 13, no. 8, 2022, ISSN: 2078-2489.
@article{info13080395,
title = {Federated Learning of Explainable AI Models in 6G Systems: Towards Secure and Automated Vehicle Networking},
author = {Alessandro Renda and Pietro Ducange and Francesco Marcelloni and Dario Sabella and Miltiadis C. Filippou and Giovanni Nardini and Giovanni Stea and Antonio Virdis and Davide Micheli and Damiano Rapone and Leonardo Gomes Baltar},
url = {https://www.mdpi.com/2078-2489/13/8/395},
doi = {10.3390/info13080395},
issn = {2078-2489},
year = {2022},
date = {2022-01-01},
urldate = {2022-01-01},
journal = {Information},
volume = {13},
number = {8},
abstract = {This article presents the concept of federated learning (FL) of eXplainable Artificial Intelligence (XAI) models as an enabling technology in advanced 5G towards 6G systems and discusses its applicability to the automated vehicle networking use case. Although the FL of neural networks has been widely investigated exploiting variants of stochastic gradient descent as the optimization method, it has not yet been adequately studied in the context of inherently explainable models. On the one side, XAI permits improving user experience of the offered communication services by helping end users trust (by design) that in-network AI functionality issues appropriate action recommendations. On the other side, FL ensures security and privacy of both vehicular and user data across the whole system. These desiderata are often ignored in existing AI-based solutions for wireless network planning, design and operation. In this perspective, the article provides a detailed description of relevant 6G use cases, with a focus on vehicle-to-everything (V2X) environments: we describe a framework to evaluate the proposed approach involving online training based on real data from live networks. FL of XAI models is expected to bring benefits as a methodology for achieving seamless availability of decentralized, lightweight and communication efficient intelligence. Impacts of the proposed approach (including standardization perspectives) consist in a better trustworthiness of operations, e.g., via explainability of quality of experience (QoE) predictions, along with security and privacy-preserving management of data from sensors, terminals, users and applications.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This article presents the concept of federated learning (FL) of eXplainable Artificial Intelligence (XAI) models as an enabling technology in advanced 5G towards 6G systems and discusses its applicability to the automated vehicle networking use case. Although the FL of neural networks has been widely investigated exploiting variants of stochastic gradient descent as the optimization method, it has not yet been adequately studied in the context of inherently explainable models. On the one side, XAI permits improving user experience of the offered communication services by helping end users trust (by design) that in-network AI functionality issues appropriate action recommendations. On the other side, FL ensures security and privacy of both vehicular and user data across the whole system. These desiderata are often ignored in existing AI-based solutions for wireless network planning, design and operation. In this perspective, the article provides a detailed description of relevant 6G use cases, with a focus on vehicle-to-everything (V2X) environments: we describe a framework to evaluate the proposed approach involving online training based on real data from live networks. FL of XAI models is expected to bring benefits as a methodology for achieving seamless availability of decentralized, lightweight and communication efficient intelligence. Impacts of the proposed approach (including standardization perspectives) consist in a better trustworthiness of operations, e.g., via explainability of quality of experience (QoE) predictions, along with security and privacy-preserving management of data from sensors, terminals, users and applications.
Fiaschi, Lorenzo; Cococcioni, Marco
A Non-Archimedean Interior Point Method and Its Application to the Lexicographic Multi-Objective Quadratic Programming Journal Article
In: Mathematics, vol. 10, no. 23, 2022, ISSN: 2227-7390.
@article{math10234536,
title = {A Non-Archimedean Interior Point Method and Its Application to the Lexicographic Multi-Objective Quadratic Programming},
author = {Lorenzo Fiaschi and Marco Cococcioni},
url = {https://www.mdpi.com/2227-7390/10/23/4536},
doi = {10.3390/math10234536},
issn = {2227-7390},
year = {2022},
date = {2022-01-01},
journal = {Mathematics},
volume = {10},
number = {23},
abstract = {This work presents a generalized implementation of the infeasible primal-dual interior point method (IPM) achieved by the use of non-Archimedean values, i.e., infinite and infinitesimal numbers. The extended version, called here the non-Archimedean IPM (NA-IPM), is proved to converge in polynomial time to a global optimum and to be able to manage infeasibility and unboundedness transparently, i.e., without considering them as corner cases: by means of a mild embedding (addition of two variables and one constraint), the NA-IPM implicitly and transparently manages their possible presence. Moreover, the new algorithm is able to solve a wider variety of linear and quadratic optimization problems than its standard counterpart. Among them, the lexicographic multi-objective one deserves particular attention, since the NA-IPM overcomes the issues that standard techniques (such as scalarization or preemptive approach) have. To support the theoretical properties of the NA-IPM, the manuscript also shows four linear and quadratic non-Archimedean programming test cases where the effectiveness of the algorithm is verified. This also stresses that the NA-IPM is not just a mere symbolic or theoretical algorithm but actually a concrete numerical tool, paving the way for its use in real-world problems in the near future.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
This work presents a generalized implementation of the infeasible primal-dual interior point method (IPM) achieved by the use of non-Archimedean values, i.e., infinite and infinitesimal numbers. The extended version, called here the non-Archimedean IPM (NA-IPM), is proved to converge in polynomial time to a global optimum and to be able to manage infeasibility and unboundedness transparently, i.e., without considering them as corner cases: by means of a mild embedding (addition of two variables and one constraint), the NA-IPM implicitly and transparently manages their possible presence. Moreover, the new algorithm is able to solve a wider variety of linear and quadratic optimization problems than its standard counterpart. Among them, the lexicographic multi-objective one deserves particular attention, since the NA-IPM overcomes the issues that standard techniques (such as scalarization or preemptive approach) have. To support the theoretical properties of the NA-IPM, the manuscript also shows four linear and quadratic non-Archimedean programming test cases where the effectiveness of the algorithm is verified. This also stresses that the NA-IPM is not just a mere symbolic or theoretical algorithm but actually a concrete numerical tool, paving the way for its use in real-world problems in the near future.
Emilio, Paolini; Marinis, Lorenzo De; Cococcioni, Marco; Valcarenghi, Luca; Maggiani, Luca; Andriolli, Nicola; others,
Photonic-Aware Neural Networks Journal Article
In: NEURAL COMPUTING & APPLICATIONS, 2022.
@article{emilio2022photonic,
title = {Photonic-Aware Neural Networks},
author = {Paolini Emilio and Lorenzo De Marinis and Marco Cococcioni and Luca Valcarenghi and Luca Maggiani and Nicola Andriolli and others},
doi = {https://doi.org/10.1007/s00521-022-07243-z},
year = {2022},
date = {2022-01-01},
journal = {NEURAL COMPUTING & APPLICATIONS},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Passaro, Lucia C.; Bondielli, Alessandro; Dell’Oglio, Pietro; Lenci, Alessandro; Marcelloni, Francesco
In-context annotation of topic-oriented datasets of fake news: A case study on the notre-dame fire event Journal Article
In: Information Sciences, vol. 615, pp. 657-677, 2022, ISSN: 0020-0255.
@article{PASSARO2022657,
title = {In-context annotation of topic-oriented datasets of fake news: A case study on the notre-dame fire event},
author = {Lucia C. Passaro and Alessandro Bondielli and Pietro Dell’Oglio and Alessandro Lenci and Francesco Marcelloni},
url = {https://www.sciencedirect.com/science/article/pii/S0020025522008167},
doi = {https://doi.org/10.1016/j.ins.2022.07.128},
issn = {0020-0255},
year = {2022},
date = {2022-01-01},
journal = {Information Sciences},
volume = {615},
pages = {657-677},
abstract = {The problem of fake news detection is becoming increasingly interesting for several research fields. Different approaches have been proposed, based on either the content of the news itself or the context and properties of its spread over time, specifically on social media. In the literature, it does not exist a widely accepted general-purpose dataset for fake news detection, due to the complexity of the task and the increasing ability to produce fake news appearing credible in particular moments. In this paper, we propose a methodology to collect and label news pertinent to specific topics and subjects. Our methodology focuses on collecting data from social media about real-world events which are known to trigger fake news. We propose a labelling method based on crowdsourcing that is fast, reliable, and able to approximate expert human annotation. The proposed method exploits both the content of the data (i.e., the texts) and contextual information about fake news for a particular real-world event. The methodology is applied to collect and annotate the Notre-Dame Fire Dataset and to annotate part of the PHEME dataset. Evaluation is performed with fake news classifiers based on Transformers and fine-tuning. Results show that context-based annotation outperforms traditional crowdsourcing out-of-context annotation.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The problem of fake news detection is becoming increasingly interesting for several research fields. Different approaches have been proposed, based on either the content of the news itself or the context and properties of its spread over time, specifically on social media. In the literature, it does not exist a widely accepted general-purpose dataset for fake news detection, due to the complexity of the task and the increasing ability to produce fake news appearing credible in particular moments. In this paper, we propose a methodology to collect and label news pertinent to specific topics and subjects. Our methodology focuses on collecting data from social media about real-world events which are known to trigger fake news. We propose a labelling method based on crowdsourcing that is fast, reliable, and able to approximate expert human annotation. The proposed method exploits both the content of the data (i.e., the texts) and contextual information about fake news for a particular real-world event. The methodology is applied to collect and annotate the Notre-Dame Fire Dataset and to annotate part of the PHEME dataset. Evaluation is performed with fake news classifiers based on Transformers and fine-tuning. Results show that context-based annotation outperforms traditional crowdsourcing out-of-context annotation.
Gallo, Gionatan; Rienzo, Francesco Di; Garzelli, Federico; Ducange, Pietro; Vallati, Carlo
A Smart System for Personal Protective Equipment Detection in Industrial Environments Based on Deep Learning at the Edge Journal Article
In: IEEE Access, vol. 10, pp. 110862-110878, 2022.
@article{9925236,
title = {A Smart System for Personal Protective Equipment Detection in Industrial Environments Based on Deep Learning at the Edge},
author = {Gionatan Gallo and Francesco Di Rienzo and Federico Garzelli and Pietro Ducange and Carlo Vallati},
doi = {10.1109/ACCESS.2022.3215148},
year = {2022},
date = {2022-01-01},
journal = {IEEE Access},
volume = {10},
pages = {110862-110878},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Velez-Estevez, A.; Ducange, P.; Perez, I. J.; Cobo, M. J.
Conceptual structure of federated learning research field Journal Article
In: Procedia Computer Science, vol. 214, pp. 1374-1381, 2022, ISSN: 1877-0509, (9th International Conference on Information Technology and Quantitative Management).
@article{VELEZESTEVEZ20221374,
title = {Conceptual structure of federated learning research field},
author = {A. Velez-Estevez and P. Ducange and I. J. Perez and M. J. Cobo},
url = {https://www.sciencedirect.com/science/article/pii/S1877050922020312},
doi = {https://doi.org/10.1016/j.procs.2022.11.319},
issn = {1877-0509},
year = {2022},
date = {2022-01-01},
journal = {Procedia Computer Science},
volume = {214},
pages = {1374-1381},
abstract = {Nowadays there are a great amount of data that can be used to train artificial intelligent systems for classification, or prediction purposes. Although there are tons of publicly available data, there are also very valuable data that is private, and therefore, it can not be shared without breaking the data protections laws. For example, hospital data has great value, but it involves persons, so we must try to preserve their privacy rights. Furthermore, although it could be interesting to train a model with the data of only one entity (i.e. a hospital), it could have more value to train the model with the data of several entities. But, since the data of each entity might not be shared, it is not possible to train a global model. In that sense, Federated Learning has emerged as a research field that deals with the training of complex models, without the necessity to share data, and therefore, keeping the data private. In this contribution, we present a global conceptual analysis based on co-words networks of the Federated Learning research field. To do that, the field was delimited using an advance query in Web of Science. The corpus contain a total of 2444 documents. As the main result, it should be highlighted that the Federated Learning research field is focused on six main global areas: telecommunications, privacy and security, computer architecture and data modeling, machine learning, and applications.},
note = {9th International Conference on Information Technology and Quantitative Management},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Nowadays there are a great amount of data that can be used to train artificial intelligent systems for classification, or prediction purposes. Although there are tons of publicly available data, there are also very valuable data that is private, and therefore, it can not be shared without breaking the data protections laws. For example, hospital data has great value, but it involves persons, so we must try to preserve their privacy rights. Furthermore, although it could be interesting to train a model with the data of only one entity (i.e. a hospital), it could have more value to train the model with the data of several entities. But, since the data of each entity might not be shared, it is not possible to train a global model. In that sense, Federated Learning has emerged as a research field that deals with the training of complex models, without the necessity to share data, and therefore, keeping the data private. In this contribution, we present a global conceptual analysis based on co-words networks of the Federated Learning research field. To do that, the field was delimited using an advance query in Web of Science. The corpus contain a total of 2444 documents. As the main result, it should be highlighted that the Federated Learning research field is focused on six main global areas: telecommunications, privacy and security, computer architecture and data modeling, machine learning, and applications.
Lai, Leonardo; Fiaschi, Lorenzo; Cococcioni, Marco; Deb, Kalyanmoy
On the Use of Grossone Methodology for Handling Priorities in Multi-objective Evolutionary Optimization Book Chapter
In: Sergeyev, Yaroslav D.; Leone, Renato De (Ed.): Numerical Infinities and Infinitesimals in Optimization, pp. 183–218, Springer International Publishing, Cham, 2022, ISBN: 978-3-030-93642-6.
@inbook{Lai2022,
title = {On the Use of Grossone Methodology for Handling Priorities in Multi-objective Evolutionary Optimization},
author = {Leonardo Lai and Lorenzo Fiaschi and Marco Cococcioni and Kalyanmoy Deb},
editor = {Yaroslav D. Sergeyev and Renato De Leone},
url = {https://doi.org/10.1007/978-3-030-93642-6_8},
doi = {10.1007/978-3-030-93642-6_8},
isbn = {978-3-030-93642-6},
year = {2022},
date = {2022-01-01},
booktitle = {Numerical Infinities and Infinitesimals in Optimization},
pages = {183–218},
publisher = {Springer International Publishing},
address = {Cham},
abstract = {This chapter introduces a new class of optimization problems, called Mixed Pareto-Lexicographic Multi-objective Optimization Problems (MPL-MOPs), to provide a suitable model for scenarios where some objectives have priority over some others. Specifically, this work focuses on two relevant subclasses of MPL-MOPs, namely optimization problems having the objective functions organized as priority chains or priority levels. A priority chain (PC) is a sequence of objectives ordered lexicographically by importance; conversely, a priority level (PL) is a group of objectives having the same importance in terms of optimization, but a lexicographic ordering exists between the PLs. After describing these problems and discussing why the standard algorithms are inadequate, an innovative approach to deal with them is introduced: it leverages the Grossone Methodology, a recent theory that allows handling priorities by means of infinite and infinitesimal numbers. Most interestingly, this technique can be easily embedded in most of the existing evolutionary algorithms, without altering their core logic. Three algorithms for MPL-MOPs are shown: the first two, called PC-NSGA-II and PC-MOEA/D, are the generalization of NSGA-II and MOEA/D, respectively, in the presence of PCs; the third, named PL-NSGA-II, generalizes instead NSGA-II when PLs are present. Several benchmark problems, including some from the real world, are used to evaluate the effectiveness of the proposed approach. The generalized algorithms are compared to other famous evolutionary ones, either priority-based or not, through a statistical analysis of their performances. The experiments show that the generalized algorithms are consistently able to produce more solutions and of higher quality.},
keywords = {},
pubstate = {published},
tppubtype = {inbook}
}
This chapter introduces a new class of optimization problems, called Mixed Pareto-Lexicographic Multi-objective Optimization Problems (MPL-MOPs), to provide a suitable model for scenarios where some objectives have priority over some others. Specifically, this work focuses on two relevant subclasses of MPL-MOPs, namely optimization problems having the objective functions organized as priority chains or priority levels. A priority chain (PC) is a sequence of objectives ordered lexicographically by importance; conversely, a priority level (PL) is a group of objectives having the same importance in terms of optimization, but a lexicographic ordering exists between the PLs. After describing these problems and discussing why the standard algorithms are inadequate, an innovative approach to deal with them is introduced: it leverages the Grossone Methodology, a recent theory that allows handling priorities by means of infinite and infinitesimal numbers. Most interestingly, this technique can be easily embedded in most of the existing evolutionary algorithms, without altering their core logic. Three algorithms for MPL-MOPs are shown: the first two, called PC-NSGA-II and PC-MOEA/D, are the generalization of NSGA-II and MOEA/D, respectively, in the presence of PCs; the third, named PL-NSGA-II, generalizes instead NSGA-II when PLs are present. Several benchmark problems, including some from the real world, are used to evaluate the effectiveness of the proposed approach. The generalized algorithms are compared to other famous evolutionary ones, either priority-based or not, through a statistical analysis of their performances. The experiments show that the generalized algorithms are consistently able to produce more solutions and of higher quality.
Cococcioni, Marco; Fiaschi, Lorenzo; Lambertini, Luca
Computing Optimal Decision Strategies Using the Infinity Computer: The Case of Non-Archimedean Zero-Sum Games Book Chapter
In: Sergeyev, Yaroslav D.; Leone, Renato De (Ed.): Numerical Infinities and Infinitesimals in Optimization, pp. 271–295, Springer International Publishing, Cham, 2022, ISBN: 978-3-030-93642-6.
@inbook{Cococcioni2022,
title = {Computing Optimal Decision Strategies Using the Infinity Computer: The Case of Non-Archimedean Zero-Sum Games},
author = {Marco Cococcioni and Lorenzo Fiaschi and Luca Lambertini},
editor = {Yaroslav D. Sergeyev and Renato De Leone},
url = {https://doi.org/10.1007/978-3-030-93642-6_11},
doi = {10.1007/978-3-030-93642-6_11},
isbn = {978-3-030-93642-6},
year = {2022},
date = {2022-01-01},
booktitle = {Numerical Infinities and Infinitesimals in Optimization},
pages = {271–295},
publisher = {Springer International Publishing},
address = {Cham},
abstract = {As is well known, zero-sum games are appropriate instruments for the analysis of several issues across areas including economics, international relations and engineering, among others. In particular, the Nash equilibria of any two-player finite zero-sum game in mixed-strategies can be found solving a proper linear programming problem. This chapter investigates and solves non-Archimedean zero-sum games, i.e., games satisfying the zero-sum property allowing the payoffs to be infinite, finite and infinitesimal. Since any zero-sum game is coupled with a linear programming problem, the search for Nash equilibria of non-Archimedean games requires the optimization of a non-Archimedean linear programming problem whose peculiarity is to have the constraints matrix populated by both infinite and infinitesimal numbers. This fact leads to the implementation of a novel non-Archimedean version of the Simplex algorithm called Gross-Matrix-Simplex. Four numerical experiments served as test cases to verify the effectiveness and correctness of the new algorithm. Moreover, these studies helped in stressing the difference between numerical and symbolic calculations: indeed, the solution output by the Gross-Matrix Simplex is just an approximation of the true Nash equilibrium, but it still satisfies some properties which resemble the idea of a non-Archimedean $$backslashvarepsilon $$ε-Nash equilibrium. On the contrary, symbolic tools seem to be able to compute the ``exact'' solution, a fact which happens only on very simple benchmarks and at the price of its intelligibility. In the general case, nevertheless, they stuck as soon as the problem becomes a little more challenging, ending up to be of little help in practice, such as in real time computations. Some possible applications related to such non-Archimedean zero-sum games are also discussed.},
keywords = {},
pubstate = {published},
tppubtype = {inbook}
}
As is well known, zero-sum games are appropriate instruments for the analysis of several issues across areas including economics, international relations and engineering, among others. In particular, the Nash equilibria of any two-player finite zero-sum game in mixed-strategies can be found solving a proper linear programming problem. This chapter investigates and solves non-Archimedean zero-sum games, i.e., games satisfying the zero-sum property allowing the payoffs to be infinite, finite and infinitesimal. Since any zero-sum game is coupled with a linear programming problem, the search for Nash equilibria of non-Archimedean games requires the optimization of a non-Archimedean linear programming problem whose peculiarity is to have the constraints matrix populated by both infinite and infinitesimal numbers. This fact leads to the implementation of a novel non-Archimedean version of the Simplex algorithm called Gross-Matrix-Simplex. Four numerical experiments served as test cases to verify the effectiveness and correctness of the new algorithm. Moreover, these studies helped in stressing the difference between numerical and symbolic calculations: indeed, the solution output by the Gross-Matrix Simplex is just an approximation of the true Nash equilibrium, but it still satisfies some properties which resemble the idea of a non-Archimedean $$backslashvarepsilon $$ε-Nash equilibrium. On the contrary, symbolic tools seem to be able to compute the ``exact'' solution, a fact which happens only on very simple benchmarks and at the price of its intelligibility. In the general case, nevertheless, they stuck as soon as the problem becomes a little more challenging, ending up to be of little help in practice, such as in real time computations. Some possible applications related to such non-Archimedean zero-sum games are also discussed.
Cococcioni, Marco; Cudazzo, Alessandro; Pappalardo, Massimo; Sergeyev, Yaroslav D.
Multi-objective Lexicographic Mixed-Integer Linear Programming: An Infinity Computer Approach Book Chapter
In: Sergeyev, Yaroslav D.; Leone, Renato De (Ed.): Numerical Infinities and Infinitesimals in Optimization, pp. 119–149, Springer International Publishing, Cham, 2022, ISBN: 978-3-030-93642-6.
@inbook{Cococcioni2022b,
title = {Multi-objective Lexicographic Mixed-Integer Linear Programming: An Infinity Computer Approach},
author = {Marco Cococcioni and Alessandro Cudazzo and Massimo Pappalardo and Yaroslav D. Sergeyev},
editor = {Yaroslav D. Sergeyev and Renato De Leone},
url = {https://doi.org/10.1007/978-3-030-93642-6_5},
doi = {10.1007/978-3-030-93642-6_5},
isbn = {978-3-030-93642-6},
year = {2022},
date = {2022-01-01},
booktitle = {Numerical Infinities and Infinitesimals in Optimization},
pages = {119–149},
publisher = {Springer International Publishing},
address = {Cham},
abstract = {In this chapter we show how a lexicographic multi-objective linear programming problem (LMOLP) can be transformed into an equivalent, single-objective one, by using the Grossone Methodology. Then we provide a simplex-like algorithm, called GrossSimplex, able to solve the original LMOLP problem using a single run of the algorithm (its theoretical correctness is also provided). In the second part, we tackle a Mixed-Integer Lexicographic Multi-Objective Linear Programming problem (LMOMILP) and we solve it in an exact way, by using a Grossone-version of the Branch-and-Bound scheme (called GrossBB). After proving the theoretical correctness of the associated pruning rules and terminating conditions, we show a few experimental results, run on an Infinity Computer simulator.},
keywords = {},
pubstate = {published},
tppubtype = {inbook}
}
In this chapter we show how a lexicographic multi-objective linear programming problem (LMOLP) can be transformed into an equivalent, single-objective one, by using the Grossone Methodology. Then we provide a simplex-like algorithm, called GrossSimplex, able to solve the original LMOLP problem using a single run of the algorithm (its theoretical correctness is also provided). In the second part, we tackle a Mixed-Integer Lexicographic Multi-Objective Linear Programming problem (LMOMILP) and we solve it in an exact way, by using a Grossone-version of the Branch-and-Bound scheme (called GrossBB). After proving the theoretical correctness of the associated pruning rules and terminating conditions, we show a few experimental results, run on an Infinity Computer simulator.
Bárcena, José Luis Corcuera; Ducange, Pietro; Ercolani, Alessio; Marcelloni, Francesco; Renda, Alessandro
An Approach to Federated Learning of Explainable Fuzzy Regression Models Proceedings Article
In: 2022 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 1-8, 2022.
@inproceedings{9882881,
title = {An Approach to Federated Learning of Explainable Fuzzy Regression Models},
author = {José Luis Corcuera Bárcena and Pietro Ducange and Alessio Ercolani and Francesco Marcelloni and Alessandro Renda},
doi = {10.1109/FUZZ-IEEE55066.2022.9882881},
year = {2022},
date = {2022-01-01},
booktitle = {2022 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE)},
pages = {1-8},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Bechini, Alessio; Bárcena, José Luis Corcuera; Ducange, Pietro; Marcelloni, Francesco; Renda, Alessandro
Increasing Accuracy and Explainability in Fuzzy Regression Trees: An Experimental Analysis Proceedings Article
In: 2022 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), pp. 1-8, 2022.
@inproceedings{9882604,
title = {Increasing Accuracy and Explainability in Fuzzy Regression Trees: An Experimental Analysis},
author = {Alessio Bechini and José Luis Corcuera Bárcena and Pietro Ducange and Francesco Marcelloni and Alessandro Renda},
doi = {10.1109/FUZZ-IEEE55066.2022.9882604},
year = {2022},
date = {2022-01-01},
booktitle = {2022 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE)},
pages = {1-8},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Lai, Leonardo; Fiaschi, Lorenzo; Cococcioni, Marco; Deb, Kalyanmoy
Pure and Mixed Lexicographic-Paretian Many-Objective Optimization: State of the Art Journal Article
In: Natural Computing, vol. 22, 2022.
@article{article,
title = {Pure and Mixed Lexicographic-Paretian Many-Objective Optimization: State of the Art},
author = {Leonardo Lai and Lorenzo Fiaschi and Marco Cococcioni and Kalyanmoy Deb},
doi = {10.1007/s11047-022-09911-4},
year = {2022},
date = {2022-01-01},
journal = {Natural Computing},
volume = {22},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Paolini, E.; Marinis, L. De; Cococcioni, M.; Valcarenghi, L.; Maggiani, L.; Andriolli, N.
Photonic-Aware Neural Network: a fixed-point emulation of photonic hardware Proceedings Article
In: 2022 27th OptoElectronics and Communications Conference (OECC) and 2022 International Conference on Photonics in Switching and Computing (PSC), pp. 01-03, 2022.
@inproceedings{9850019,
title = {Photonic-Aware Neural Network: a fixed-point emulation of photonic hardware},
author = {E. Paolini and L. De Marinis and M. Cococcioni and L. Valcarenghi and L. Maggiani and N. Andriolli},
doi = {10.23919/OECC/PSC53152.2022.9850019},
year = {2022},
date = {2022-01-01},
booktitle = {2022 27th OptoElectronics and Communications Conference (OECC) and 2022 International Conference on Photonics in Switching and Computing (PSC)},
pages = {01-03},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}