2025
Other
Open Access
Manuale di identità visiva dell’Istituto di scienza e tecnologie dell’informazione del Cnr (Cnr-Isti)
Rapisarda B., Genovali K.
Questo manuale di identità visiva offre una guida per applicare in modo coerente ed efficace il logo e gli elementi dell'identità visiva del Cnr-Isti nei materiali di comunicazione, sia digitali che cartacei.DOI: 10.32079/isti-tr-2025/003
Metrics:
Rapisarda B., Genovali K.
Questo manuale di identità visiva offre una guida per applicare in modo coerente ed efficace il logo e gli elementi dell'identità visiva del Cnr-Isti nei materiali di comunicazione, sia digitali che cartacei.DOI: 10.32079/isti-tr-2025/003
Metrics:
2025
Conference article
Open Access
Marginal cost of computation as a collaborative strategy for resource management at the edge
Carlini E., Dazzi P., Ferrucci L., Massa J., Mordacchini M.
This paper explores an extended applications’ cost function to model the willingness of Edge data centres to accommodate additional users in decentralized edge computing environments. By enhancing the Marginal Computing Cost per User (MCU) concept, we introduce a dynamic cost factor influenced by the number of users currently served. Through extensive simulations conducted on the PureEdgeSim platform, we evaluate the impact of this variable MCU on system performance across various configurations. The results reveal a critical trade-off between cost sensitivity (i.e., collaboration willingness) of Edge data centres and optimization potential. This work offers insights into user allocation strategies in heterogeneous edge systems and sets the stage for future research into non-linear MCU configurations and diverse application workloads.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15358, pp. 15-27. Rome, Italy, 26-27/09/2024
DOI: 10.1007/978-3-031-81226-2_3
Project(s): EDGELESS
Metrics:
Carlini E., Dazzi P., Ferrucci L., Massa J., Mordacchini M.
This paper explores an extended applications’ cost function to model the willingness of Edge data centres to accommodate additional users in decentralized edge computing environments. By enhancing the Marginal Computing Cost per User (MCU) concept, we introduce a dynamic cost factor influenced by the number of users currently served. Through extensive simulations conducted on the PureEdgeSim platform, we evaluate the impact of this variable MCU on system performance across various configurations. The results reveal a critical trade-off between cost sensitivity (i.e., collaboration willingness) of Edge data centres and optimization potential. This work offers insights into user allocation strategies in heterogeneous edge systems and sets the stage for future research into non-linear MCU configurations and diverse application workloads.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15358, pp. 15-27. Rome, Italy, 26-27/09/2024
DOI: 10.1007/978-3-031-81226-2_3
Project(s): EDGELESS
Metrics:
See at:
CNR IRIS 
| link.springer.com | CNR IRIS 
| CNR IRIS
2025
Conference article
Restricted
Maybe you are looking for CroQS Cross-Modal Query Suggestion for text-to-image retrieval
Pacini G., Carrara F., Messina N., Tonellotto N., Amato G., Falchi F.
Query suggestion, a technique widely adopted in information retrieval, enhances system interactivity and the browsing experience of document collections. In cross-modal retrieval, many works have focused on retrieving relevant items from natural language queries, while few have explored query suggestion solutions. In this work, we address query suggestion in cross-modal retrieval, introducing a novel task that focuses on suggesting minimal textual modifications needed to explore visually consistent subsets of the collection, following the premise of “Maybe you are looking for”. To facilitate the evaluation and development of methods, we present a tailored benchmark named CroQS. This dataset comprises initial queries, grouped result sets, and human-defined suggested queries for each group. We establish dedicated metrics to rigorously evaluate the performance of various methods on this task, measuring representativeness, cluster specificity, and similarity of the suggested queries to the original ones. Baseline methods from related fields, such as image captioning and content summarization, are adapted for this task to provide reference performance scores. Although relatively far from human performance, our experiments reveal that both LLM-based and captioning-based methods achieve competitive results on CroQS, improving the recall on cluster specificity by more than 115% and representativeness mAP by more than 52% with respect to the initial query. The dataset, the implementation of the baseline methods and the notebooks containing our experiments are available here: paciosoft.com/CroQS-benchmark/.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15573, pp. 138-152. Lucca, Italy, April 6–10, 2025
DOI: 10.1007/978-3-031-88711-6_9
Project(s): Future Artificial Intelligence Research, a MUltimedia platform for Content Enrichment and Search in audiovisual archives
Metrics:
Pacini G., Carrara F., Messina N., Tonellotto N., Amato G., Falchi F.
Query suggestion, a technique widely adopted in information retrieval, enhances system interactivity and the browsing experience of document collections. In cross-modal retrieval, many works have focused on retrieving relevant items from natural language queries, while few have explored query suggestion solutions. In this work, we address query suggestion in cross-modal retrieval, introducing a novel task that focuses on suggesting minimal textual modifications needed to explore visually consistent subsets of the collection, following the premise of “Maybe you are looking for”. To facilitate the evaluation and development of methods, we present a tailored benchmark named CroQS. This dataset comprises initial queries, grouped result sets, and human-defined suggested queries for each group. We establish dedicated metrics to rigorously evaluate the performance of various methods on this task, measuring representativeness, cluster specificity, and similarity of the suggested queries to the original ones. Baseline methods from related fields, such as image captioning and content summarization, are adapted for this task to provide reference performance scores. Although relatively far from human performance, our experiments reveal that both LLM-based and captioning-based methods achieve competitive results on CroQS, improving the recall on cluster specificity by more than 115% and representativeness mAP by more than 52% with respect to the initial query. The dataset, the implementation of the baseline methods and the notebooks containing our experiments are available here: paciosoft.com/CroQS-benchmark/.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15573, pp. 138-152. Lucca, Italy, April 6–10, 2025
DOI: 10.1007/978-3-031-88711-6_9
Project(s): Future Artificial Intelligence Research, a MUltimedia platform for Content Enrichment and Search in audiovisual archives
Metrics:
See at:
CNR IRIS | CNR IRIS | CNR IRIS | link.springer.com
2025
Other
Open Access
Piano di comunicazione dell'Istituto di scienza e tecnologie dell'informazione "A. Faedo" 2024-2025
Rapisarda B., Genovali K., Falconetti L.
Il piano di comunicazione contiene la strategia di comunicazione che l'Istituto intende adottare nel periodo di riferimento indicato e la descrizione di tutte le azioni che si è deciso di intraprendere per raggiungere gli obiettivi di comunicazione preposti. Con l’adozione del piano ci si propone di: mettere ordine nello svolgimento delle attività di comunicazione già in corso, definendo e assegnando ruoli specifici sia al personale referente che per la comunicazione che al personale di ricerca, così come stabilendo procedure nuove o aggiuntive di passaggio delle informazioni; costruire un piano editoriale della comunicazione online codificato e stabile; fornire informazioni e procedure chiare ed efficaci al personale di ricerca, per renderlo il più possibile autonomo nelle proprie attività di comunicazione.DOI: 10.32079/isti-tr-2025/002
Metrics:
Rapisarda B., Genovali K., Falconetti L.
Il piano di comunicazione contiene la strategia di comunicazione che l'Istituto intende adottare nel periodo di riferimento indicato e la descrizione di tutte le azioni che si è deciso di intraprendere per raggiungere gli obiettivi di comunicazione preposti. Con l’adozione del piano ci si propone di: mettere ordine nello svolgimento delle attività di comunicazione già in corso, definendo e assegnando ruoli specifici sia al personale referente che per la comunicazione che al personale di ricerca, così come stabilendo procedure nuove o aggiuntive di passaggio delle informazioni; costruire un piano editoriale della comunicazione online codificato e stabile; fornire informazioni e procedure chiare ed efficaci al personale di ricerca, per renderlo il più possibile autonomo nelle proprie attività di comunicazione.DOI: 10.32079/isti-tr-2025/002
Metrics:
2025
Other
Open Access
Come fare per... Indicazioni per una comunicazione più efficace in Isti
Rapisarda B., Genovali K.
Questo documento fornisce le indicazioni essenziali per orientare i ricercatori e tutto il personale dell'Istituto verso una comunicazione istituzionale efficace. Al suo interno troverete dettagli che potranno supportarvi sia nella definizione di una strategia comunicativa di base che nella realizzazione di piccoli ma accurati prodotti di comunicazione.DOI: 10.32079/isti-tr-2025/004
Metrics:
Rapisarda B., Genovali K.
Questo documento fornisce le indicazioni essenziali per orientare i ricercatori e tutto il personale dell'Istituto verso una comunicazione istituzionale efficace. Al suo interno troverete dettagli che potranno supportarvi sia nella definizione di una strategia comunicativa di base che nella realizzazione di piccoli ma accurati prodotti di comunicazione.DOI: 10.32079/isti-tr-2025/004
Metrics:
2025
Journal article
Open Access
The open-closed mod-minimizer algorithm
Koerkamp R. G., Liu D., Pibiri G. E.
Sampling algorithms that deterministically select a subset of k-mers are an important building block in bioinformatics applications. For example, they are used to index large textual collections, like DNA, and to compare sequences quickly. In such applications, a sampling algorithm is required to select one k-mer out of every window of w consecutive k-mers. The folklore and most used scheme is the random minimizer that selects the smallest k-mer in the window according to some random order. This scheme is remarkably simple and versatile, and has a density (expected fraction of selected k-mers) of 2/(w+1). In practice, lower density leads to faster methods and smaller indexes, and it turns out that the random minimizer is not the best one can do. Indeed, some schemes are known to approach optimal density 1/w when k→∞, like the recently introduced mod-minimizer (Groot Koerkamp and Pibiri, WABI 2024). In this work, we study methods that achieve low density when k≤w. In this small-k regime, a practical method with provably better density than the random minimizer is the miniception (Zheng et al., Bioinformatics 2021). This method can be elegantly described as sampling the smallest closed sycnmer (Edgar, PeerJ 2021) in the window according to some random order. We show that extending the miniception to prefer sampling open syncmers yields much better density. This new method—the open-closed minimizer—offers improved density for small k≤w while being as fast to compute as the random minimizer. Compared to methods based on decycling sets, that achieve very low density in the small-k regime, our method has comparable density while being computationally simpler and intuitive. Furthermore, we extend the mod-minimizer to improve density of any scheme that works well for small k to also work well when k>w is large. We hence obtain the open-closed mod-minimizer, a practical method that improves over the mod-minimizer for all k.Source: ALGORITHMS FOR MOLECULAR BIOLOGY, vol. 20 (issue 1)
DOI: 10.1186/s13015-025-00270-0
Project(s): ETH Research Grant ETH-1721, EFRA
Metrics:
Koerkamp R. G., Liu D., Pibiri G. E.
Sampling algorithms that deterministically select a subset of k-mers are an important building block in bioinformatics applications. For example, they are used to index large textual collections, like DNA, and to compare sequences quickly. In such applications, a sampling algorithm is required to select one k-mer out of every window of w consecutive k-mers. The folklore and most used scheme is the random minimizer that selects the smallest k-mer in the window according to some random order. This scheme is remarkably simple and versatile, and has a density (expected fraction of selected k-mers) of 2/(w+1). In practice, lower density leads to faster methods and smaller indexes, and it turns out that the random minimizer is not the best one can do. Indeed, some schemes are known to approach optimal density 1/w when k→∞, like the recently introduced mod-minimizer (Groot Koerkamp and Pibiri, WABI 2024). In this work, we study methods that achieve low density when k≤w. In this small-k regime, a practical method with provably better density than the random minimizer is the miniception (Zheng et al., Bioinformatics 2021). This method can be elegantly described as sampling the smallest closed sycnmer (Edgar, PeerJ 2021) in the window according to some random order. We show that extending the miniception to prefer sampling open syncmers yields much better density. This new method—the open-closed minimizer—offers improved density for small k≤w while being as fast to compute as the random minimizer. Compared to methods based on decycling sets, that achieve very low density in the small-k regime, our method has comparable density while being computationally simpler and intuitive. Furthermore, we extend the mod-minimizer to improve density of any scheme that works well for small k to also work well when k>w is large. We hence obtain the open-closed mod-minimizer, a practical method that improves over the mod-minimizer for all k.Source: ALGORITHMS FOR MOLECULAR BIOLOGY, vol. 20 (issue 1)
DOI: 10.1186/s13015-025-00270-0
Project(s): ETH Research Grant ETH-1721, EFRA
Metrics:
See at:
almob.biomedcentral.com | CNR IRIS | CNR IRIS
2024
Journal article
Open Access
Fulgor: a fast and compact k-mer index for large-scale matching and color queries
Fan J, Khan J, Pratap Singh N, Pibiri Ge, Patro R
The problem of sequence identification or matching--determining the subset of reference sequences from a givencollection that are likely to contain a short, queried nucleotide sequence--is relevant for many important tasksin Computational Biology, such as metagenomics and pangenome analysis. Due to the complex nature of such analysesand the large scale of the reference collections a resource-efficient solution to this problem is of utmost importance.This poses the threefold challenge of representing the reference collection with a data structure that is efficientto query, has light memory usage, and scales well to large collections. To solve this problem, we describe an efficientcolored de Bruijn graph index, arising as the combination of a k-mer dictionary with a compressed inverted index. Theproposed index takes full advantage of the fact that unitigs in the colored compacted de Bruijn graph are monochromatic(i.e., all k-mers in a unitig have the same set of references of origin, or color). Specifically, the unitigs are keptin the dictionary in color order, thereby allowing for the encoding of the map from k-mers to their colors in as littleas 1 + o(1) bits per unitig. Hence, one color per unitig is stored in the index with almost no space/time overhead. Bycombining this property with simple but effective compression methods for integer lists, the index achieves verysmall space. We implement these methods in a tool called Fulgor, and conduct an extensive experimental analysisto demonstrate the improvement of our tool over previous solutions. For example, compared to Themisto--thestrongest competitor in terms of index space vs. query time trade-off--Fulgor requires significantly less space (upto 43% less space for a collection of 150,000 Salmonella enterica genomes), is at least twice as fast for color queries,and is 2-6× faster to construct.Source: ALGORITHMS FOR MOLECULAR BIOLOGY, vol. 19 (issue 3)
DOI: 10.1186/s13015-024-00251-9
Metrics:
Fan J, Khan J, Pratap Singh N, Pibiri Ge, Patro R
The problem of sequence identification or matching--determining the subset of reference sequences from a givencollection that are likely to contain a short, queried nucleotide sequence--is relevant for many important tasksin Computational Biology, such as metagenomics and pangenome analysis. Due to the complex nature of such analysesand the large scale of the reference collections a resource-efficient solution to this problem is of utmost importance.This poses the threefold challenge of representing the reference collection with a data structure that is efficientto query, has light memory usage, and scales well to large collections. To solve this problem, we describe an efficientcolored de Bruijn graph index, arising as the combination of a k-mer dictionary with a compressed inverted index. Theproposed index takes full advantage of the fact that unitigs in the colored compacted de Bruijn graph are monochromatic(i.e., all k-mers in a unitig have the same set of references of origin, or color). Specifically, the unitigs are keptin the dictionary in color order, thereby allowing for the encoding of the map from k-mers to their colors in as littleas 1 + o(1) bits per unitig. Hence, one color per unitig is stored in the index with almost no space/time overhead. Bycombining this property with simple but effective compression methods for integer lists, the index achieves verysmall space. We implement these methods in a tool called Fulgor, and conduct an extensive experimental analysisto demonstrate the improvement of our tool over previous solutions. For example, compared to Themisto--thestrongest competitor in terms of index space vs. query time trade-off--Fulgor requires significantly less space (upto 43% less space for a collection of 150,000 Salmonella enterica genomes), is at least twice as fast for color queries,and is 2-6× faster to construct.Source: ALGORITHMS FOR MOLECULAR BIOLOGY, vol. 19 (issue 3)
DOI: 10.1186/s13015-024-00251-9
Metrics:
See at:
almob.biomedcentral.com | CNR IRIS | ISTI Repository | CNR IRIS
2024
Conference article
Open Access
Information dissimilarity measures in decentralized knowledge distillation: a comparative analysis
Molo M. B., Vadicamo L., Carlini E., Gennaro C., Connor R.
Knowledge distillation (KD) is a key technique for transferring knowledge from a large, complex “teacher” model to a smaller, more efficient “student” model. Although initially developed for model compression, it has found applications across various domains due to the benefits of its knowledge transfer mechanism. While Cross Entropy (CE) and Kullback-Leibler (KL) are commonly used in KD, this work investigates the applicability of loss functions based on underexplored information dissimilarity measures, such as Triangular Divergence (TD), Structural Entropic Distance (SED), and Jensen-Shannon Divergence (JS), for both independent and identically distributed (iid) and non-iid data distributions. The primary contributions of this study include an empirical evaluation of these dissimilarity measures within a decentralized learning context, i.e., where independent clients collaborate without a central server coordinating the learning process. Additionally, the paper assesses the performance of clients by comparing pairwise distillation averaging among clients to conventional peer-to-peer pairwise distillation. Results indicate that while dissimilarity measures perform comparably in iid settings, non-iid distributions favor SED and JS, which also demonstrated consistent performance across clients.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15268, pp. 140-154. Providence, USA, 4-6/11/2024
DOI: 10.1007/978-3-031-75823-2_12
Project(s): National Centre for HPC, Big Data and Quantum Computing, SUN
Metrics:
Molo M. B., Vadicamo L., Carlini E., Gennaro C., Connor R.
Knowledge distillation (KD) is a key technique for transferring knowledge from a large, complex “teacher” model to a smaller, more efficient “student” model. Although initially developed for model compression, it has found applications across various domains due to the benefits of its knowledge transfer mechanism. While Cross Entropy (CE) and Kullback-Leibler (KL) are commonly used in KD, this work investigates the applicability of loss functions based on underexplored information dissimilarity measures, such as Triangular Divergence (TD), Structural Entropic Distance (SED), and Jensen-Shannon Divergence (JS), for both independent and identically distributed (iid) and non-iid data distributions. The primary contributions of this study include an empirical evaluation of these dissimilarity measures within a decentralized learning context, i.e., where independent clients collaborate without a central server coordinating the learning process. Additionally, the paper assesses the performance of clients by comparing pairwise distillation averaging among clients to conventional peer-to-peer pairwise distillation. Results indicate that while dissimilarity measures perform comparably in iid settings, non-iid distributions favor SED and JS, which also demonstrated consistent performance across clients.Source: LECTURE NOTES IN COMPUTER SCIENCE, vol. 15268, pp. 140-154. Providence, USA, 4-6/11/2024
DOI: 10.1007/978-3-031-75823-2_12
Project(s): National Centre for HPC, Big Data and Quantum Computing, SUN
Metrics:
See at:
IRIS Cnr | IRIS Cnr | IRIS Cnr | doi.org | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
Urgent edge computing
Dazzi P., Ferrucci L., Danelutto M., Tserpes K., Makris A., Theodoropoulos T., Massa J., Carlini E., Mordacchini M.
This position paper introduces Urgent Edge Computing (UEC) as a paradigm shift addressing the evolving demands of time-sensitive applications in distributed edge environments, in time-critical scenarios. With a focus on ultra-low latency, availability, resource management, decentralization, self-organization, and robust security, UEC aims to facilitate operations in critical scenarios such as disaster response, environmental monitoring, and smart city management. This paper outlines and discusses the key requirements, challenges, and enablers along with a conceptual architecture. The paper also outlines the potential applications of Urgent Edge Computing.DOI: 10.1145/3659994.3660315
DOI: 10.48550/arxiv.2404.13411
Project(s): NOUS, Spoke 1 “FutureHPC & BigData” and Spoke 6 “Multiscale Modelling & Engineering Applications f the Italian Research Center on High-Performance Com- puting, Big Data and Quantum Computing (ICSC)
Metrics:
Dazzi P., Ferrucci L., Danelutto M., Tserpes K., Makris A., Theodoropoulos T., Massa J., Carlini E., Mordacchini M.
This position paper introduces Urgent Edge Computing (UEC) as a paradigm shift addressing the evolving demands of time-sensitive applications in distributed edge environments, in time-critical scenarios. With a focus on ultra-low latency, availability, resource management, decentralization, self-organization, and robust security, UEC aims to facilitate operations in critical scenarios such as disaster response, environmental monitoring, and smart city management. This paper outlines and discusses the key requirements, challenges, and enablers along with a conceptual architecture. The paper also outlines the potential applications of Urgent Edge Computing.DOI: 10.1145/3659994.3660315
DOI: 10.48550/arxiv.2404.13411
Project(s): NOUS
, Spoke 1 “FutureHPC & BigData” and Spoke 6 “Multiscale Modelling & Engineering Applications f the Italian Research Center on High-Performance Com- puting, Big Data and Quantum Computing (ICSC)Metrics:
See at:
arXiv.org e-Print Archive | IRIS Cnr | Archivio della Ricerca - Università di Pisa | IRIS Cnr | Archivio della Ricerca - Università di Pisa | doi.org | Archivio della Ricerca - Università di Pisa | CNR IRIS
2024
Conference article
Open Access
EDGELESS: a software architecture for stateful FaaS at the edge
Cicconetti C., Carlini E., Hetzel R., Mortier R., Paradell A., Sauer M.
EDGELESS is a serverless platform targeting edge computing that supports widely distributed deployments using heterogeneous devices. We present its components, architecture, and programming model. Our working prototype of EDGELESS enables executing lightweight functions and is already available as open-source.DOI: 10.1145/3625549.3658817
Project(s): EDGELESS
Metrics:
Cicconetti C., Carlini E., Hetzel R., Mortier R., Paradell A., Sauer M.
EDGELESS is a serverless platform targeting edge computing that supports widely distributed deployments using heterogeneous devices. We present its components, architecture, and programming model. Our working prototype of EDGELESS enables executing lightweight functions and is already available as open-source.DOI: 10.1145/3625549.3658817
Project(s): EDGELESS
Metrics:
See at:
dl.acm.org | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
Teacher-student models for AI vision at the edge: a car parking case study
Molo M. J., Carlini E., Ciampi L., Gennaro C., Vadicamo L.
The surge of the Internet of Things has sparked a multitude of deep learning-based computer vision applications that extract relevant information from the deluge of data coming from Edge devices, such as smart cameras. Nevertheless, this promising approach introduces new obstacles, including the constraints posed by the limited computational resources on these devices and the challenges associated with the generalization capabilities of the AI-based models against novel scenarios never seen during the supervised training, a situation frequently encountered in this context. This work proposes an efficient approach for detecting vehicles in parking lot scenarios monitored by multiple smart cameras that train their underlying AI-based models by exploiting knowledge distillation. Specifically, we consider an architectural scheme comprising a powerful and large detector used as a teacher and several shallow models acting as students, more appropriate for computational-bounded devices and designed to run onboard the smart cameras. The teacher is pre-trained over general-context data and behaves like an oracle, transferring its knowledge to the smaller nodes; on the other hand, the students learn to localize cars in new specific scenarios without using further labeled data, relying solely on the distilled loss coming from the oracle. Preliminary results show that student models trained only with distillation loss increase their performances, sometimes even outperforming the results achieved by the same models supervised with the ground truth.DOI: 10.5220/0012376900003660
Project(s): AI4Media, National Centre for HPC, Big Data and Quantum Computing, Sustainable Mobility Center
Metrics:
Molo M. J., Carlini E., Ciampi L., Gennaro C., Vadicamo L.
The surge of the Internet of Things has sparked a multitude of deep learning-based computer vision applications that extract relevant information from the deluge of data coming from Edge devices, such as smart cameras. Nevertheless, this promising approach introduces new obstacles, including the constraints posed by the limited computational resources on these devices and the challenges associated with the generalization capabilities of the AI-based models against novel scenarios never seen during the supervised training, a situation frequently encountered in this context. This work proposes an efficient approach for detecting vehicles in parking lot scenarios monitored by multiple smart cameras that train their underlying AI-based models by exploiting knowledge distillation. Specifically, we consider an architectural scheme comprising a powerful and large detector used as a teacher and several shallow models acting as students, more appropriate for computational-bounded devices and designed to run onboard the smart cameras. The teacher is pre-trained over general-context data and behaves like an oracle, transferring its knowledge to the smaller nodes; on the other hand, the students learn to localize cars in new specific scenarios without using further labeled data, relying solely on the distilled loss coming from the oracle. Preliminary results show that student models trained only with distillation loss increase their performances, sometimes even outperforming the results achieved by the same models supervised with the ground truth.DOI: 10.5220/0012376900003660
Project(s): AI4Media
, National Centre for HPC, Big Data and Quantum Computing, Sustainable Mobility CenterMetrics:
See at:
CNR IRIS | www.scitepress.org | CNR IRIS | CNR IRIS
2024
Conference article
Restricted
Understanding human mobility dynamics: insights from summarized semantic trajectories
Pugliese C., Lettich F., Pinelli F., Renso C.
Mobility data analysis provides insights into human movement patterns, traffic flows, and urban planning strategies. Human dynamics analysis focuses on tracking people to investigate how individuals and groups behave, interact, and evolve. Various mobility data sources, such as GPS, mobile phone records, social media, and transportation logs, are often semantically enriched and used for these analyses. This results in the generation of new, complex datasets that require effective summarization methods to reduce data volume while preserving relevant information. In this work, we aim to demonstrate the effective use of summarized semantic trajectories in analyzing human mobility behaviours. We offer empirical evidence from a case study, showing how this type of trajectory helps in understanding human mobility, especially in distinguishing between routine and non-routine behaviours. Experimental results show that the analysis results are comparable with the results obtained in the original (non summarized) dataset.DOI: 10.1109/mdm61037.2024.00039
Project(s): CAMEO, PRIN 2022 n. 2022ZLL7MW, SoBigData-PlusPlus, Spoke 1 ”Human-centered AI” of the M4C2 - Investimento 1.3, Partenariato Esteso PE00000013 - ”FAIR - Future Artificial Intelligence Research”
Metrics:
Pugliese C., Lettich F., Pinelli F., Renso C.
Mobility data analysis provides insights into human movement patterns, traffic flows, and urban planning strategies. Human dynamics analysis focuses on tracking people to investigate how individuals and groups behave, interact, and evolve. Various mobility data sources, such as GPS, mobile phone records, social media, and transportation logs, are often semantically enriched and used for these analyses. This results in the generation of new, complex datasets that require effective summarization methods to reduce data volume while preserving relevant information. In this work, we aim to demonstrate the effective use of summarized semantic trajectories in analyzing human mobility behaviours. We offer empirical evidence from a case study, showing how this type of trajectory helps in understanding human mobility, especially in distinguishing between routine and non-routine behaviours. Experimental results show that the analysis results are comparable with the results obtained in the original (non summarized) dataset.DOI: 10.1109/mdm61037.2024.00039
Project(s): CAMEO, PRIN 2022 n. 2022ZLL7MW, SoBigData-PlusPlus
, Spoke 1 ”Human-centered AI” of the M4C2 - Investimento 1.3, Partenariato Esteso PE00000013 - ”FAIR - Future Artificial Intelligence Research”Metrics:
2024
Conference article
Open Access
Calibration of continual learning models
Li L., Piccoli E., Cossu A., Bacciu D., Lomonaco V.
Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction.Source: IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, vol. 24, pp. 4160-4169. Seattle, USA, 2024
DOI: 10.1109/cvprw63382.2024.00419
Project(s): EMERGE, Future Artificial Intelligence Research
Metrics:
Li L., Piccoli E., Cossu A., Bacciu D., Lomonaco V.
Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction.Source: IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, vol. 24, pp. 4160-4169. Seattle, USA, 2024
DOI: 10.1109/cvprw63382.2024.00419
Project(s): EMERGE
, Future Artificial Intelligence ResearchMetrics:
See at:
CNR IRIS | ieeexplore.ieee.org | CNR IRIS | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
From geolocated images to urban region identification and description: a large language model approach
Rocchietti G., Pugliese C., Sartori Rangel G., Carvalho J. T.
Urban research faces challenges in understanding and describing city regions, which are essential for urban planning and tourism management. Traditional methods rely on predefined areas and non-human-readable representations. This paper presents a new unsupervised approach that overcomes these limitations using a data-driven method with Instruction-tuned Large Language Models (ILLMs). Our technique dynamically identifies urban regions with similar features and generates human-readable descriptions. We validate this method using Flickr images from Pisa, Italy, and our results show that it effectively captures the semantic features of urban regions and generates comprehensible textual descriptions.DOI: 10.1145/3678717.3691317
Metrics:
Rocchietti G., Pugliese C., Sartori Rangel G., Carvalho J. T.
Urban research faces challenges in understanding and describing city regions, which are essential for urban planning and tourism management. Traditional methods rely on predefined areas and non-human-readable representations. This paper presents a new unsupervised approach that overcomes these limitations using a data-driven method with Instruction-tuned Large Language Models (ILLMs). Our technique dynamically identifies urban regions with similar features and generates human-readable descriptions. We validate this method using Flickr images from Pisa, Italy, and our results show that it effectively captures the semantic features of urban regions and generates comprehensible textual descriptions.DOI: 10.1145/3678717.3691317
Metrics:
See at:
dl.acm.org | CNR IRIS | doi.org | CNR IRIS
2024
Journal article
Restricted
Report on the 13th Italian Information Retrieval Workshop (IIR 2023)
Faggioli G., Ferrara A., Nardini F. M., Tonellotto N.
The 13th Italian Information Retrieval Workshop is the thirteenth edition of the annual conference of the Italian information retrieval and recommender systems communities. The two days of the conference gathered interesting studies and research work on a wide range of topics on information retrieval, recommender systems, and natural language processing, such as Search and Ranking, Recommendation, Content Analysis, and Classification, Artificial Intelligence, NLP, Semantics, and Dialog, Domain-Specific Applications, Human Factors and Interfaces, and Evaluation. It was hosted by the National Research Council (CNR) of Italy and the University of Pisa in a conference facility in Pisa, Italy.Source: SIGIR FORUM, vol. 57 (issue 2), pp. 1-12
DOI: 10.1145/3642979.3642990
Metrics:
Faggioli G., Ferrara A., Nardini F. M., Tonellotto N.
The 13th Italian Information Retrieval Workshop is the thirteenth edition of the annual conference of the Italian information retrieval and recommender systems communities. The two days of the conference gathered interesting studies and research work on a wide range of topics on information retrieval, recommender systems, and natural language processing, such as Search and Ranking, Recommendation, Content Analysis, and Classification, Artificial Intelligence, NLP, Semantics, and Dialog, Domain-Specific Applications, Human Factors and Interfaces, and Evaluation. It was hosted by the National Research Council (CNR) of Italy and the University of Pisa in a conference facility in Pisa, Italy.Source: SIGIR FORUM, vol. 57 (issue 2), pp. 1-12
DOI: 10.1145/3642979.3642990
Metrics:
See at:
dl.acm.org | ACM SIGIR Forum | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
On the application of a common theoretical explainability framework in information retrieval
Albarelli A., Lucchese C., Rizzo M., Veneri A.
Most of the current state-of-the-art models used to solve the search and ranking tasks in Information Retrieval (IR) are considered “black boxes” due to the enormous number of parameters employed, which makes it difficult for humans to understand the relation between input and output. Thus, in the current literature, several approaches are proposed to explain their outputs, trying to make the models more explainable while maintaining the high level of effectiveness achieved. Even though many methods have been developed, there is still a lack of a common way of describing and evaluating the models and methods of the Explainabile IR (ExIR) field. This work shows how a common theoretical framework for explainability (previously presented in the biomedical field) can be applied to IR. We first describe the general framework and then focus on specific explanation techniques in the IR field, focusing on core IR tasks: search and ranking. We show how well-known methods in ExIR fit into the framework and how specific IR explainability evaluation metrics can be described using this new setting.Source: CEUR WORKSHOP PROCEEDINGS, vol. 3802, pp. 43-52. Udine, Italy, 05-06/09/2024
Albarelli A., Lucchese C., Rizzo M., Veneri A.
Most of the current state-of-the-art models used to solve the search and ranking tasks in Information Retrieval (IR) are considered “black boxes” due to the enormous number of parameters employed, which makes it difficult for humans to understand the relation between input and output. Thus, in the current literature, several approaches are proposed to explain their outputs, trying to make the models more explainable while maintaining the high level of effectiveness achieved. Even though many methods have been developed, there is still a lack of a common way of describing and evaluating the models and methods of the Explainabile IR (ExIR) field. This work shows how a common theoretical framework for explainability (previously presented in the biomedical field) can be applied to IR. We first describe the general framework and then focus on specific explanation techniques in the IR field, focusing on core IR tasks: search and ranking. We show how well-known methods in ExIR fit into the framework and how specific IR explainability evaluation metrics can be described using this new setting.Source: CEUR WORKSHOP PROCEEDINGS, vol. 3802, pp. 43-52. Udine, Italy, 05-06/09/2024
See at:
ceur-ws.org | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
The {mod-minimizer}: a simple and efficient sampling algorithm for long k-Mers
Koerkamp R. G., Pibiri G. E.
Motivation. Given a string S, a minimizer scheme is an algorithm defined by a triple (k,w,O) that samples a subset of k-mers (k-long substrings) from a string S. Specifically, it samples the smallest k-mer according to the order O from each window of w consecutive k-mers in S. Because consecutive windows can sample the same k-mer, the set of the sampled k-mers is typically much smaller than S. More generally, we consider substring sampling algorithms that respect a window guarantee: at least one k-mer must be sampled from every window of w consecutive k-mers. As a sampled k-mer is uniquely identified by its absolute position in S, we can define the density of a sampling algorithm as the fraction of distinct sampled positions. Good methods have low density which, by respecting the window guarantee, is lower bounded by 1/w. It is however difficult to design a sequence-agnostic algorithm with provably optimal density. In practice, the order O is usually implemented using a pseudo-random hash function to obtain the so-called random minimizer. This scheme is simple to implement, very fast to compute even in streaming fashion, and easy to analyze. However, its density is almost a factor of 2 away from the lower bound for large windows. Methods. In this work we introduce mod-sampling, a two-step sampling algorithm to obtain new minimizer schemes. Given a (small) parameter t, the mod-sampling algorithm finds the position p of the smallest t-mer in a window. It then samples the k-mer at position pod w. The lr-minimizer uses t = k-w and the mod-minimizer uses t≡ k (mod w). Results. These new schemes have provably lower density than random minimizers and other schemes when k is large compared to w, while being as fast to compute. Importantly, the mod-minimizer achieves optimal density when k → ∞. Although the mod-minimizer is not the first method to achieve optimal density for large k, its proof of optimality is simpler than previous work. We provide pseudocode for a number of other methods and compare to them. In practice, the mod-minimizer has considerably lower density than the random minimizer and other state-of-the-art methods, like closed syncmers and miniception, when k > w. We plugged the mod-minimizer into SSHash, a k-mer dictionary based on minimizers. For default parameters (w,k) = (11,21), space usage decreases by 15% when indexing the whole human genome (GRCh38), while maintaining its fast query time.Source: LEIBNIZ INTERNATIONAL PROCEEDINGS IN INFORMATICS, vol. 312. Royal Holloway, London, United Kingdom, 2-4/09/2024
DOI: 10.4230/lipics.wabi.2024.11
Metrics:
Koerkamp R. G., Pibiri G. E.
Motivation. Given a string S, a minimizer scheme is an algorithm defined by a triple (k,w,O) that samples a subset of k-mers (k-long substrings) from a string S. Specifically, it samples the smallest k-mer according to the order O from each window of w consecutive k-mers in S. Because consecutive windows can sample the same k-mer, the set of the sampled k-mers is typically much smaller than S. More generally, we consider substring sampling algorithms that respect a window guarantee: at least one k-mer must be sampled from every window of w consecutive k-mers. As a sampled k-mer is uniquely identified by its absolute position in S, we can define the density of a sampling algorithm as the fraction of distinct sampled positions. Good methods have low density which, by respecting the window guarantee, is lower bounded by 1/w. It is however difficult to design a sequence-agnostic algorithm with provably optimal density. In practice, the order O is usually implemented using a pseudo-random hash function to obtain the so-called random minimizer. This scheme is simple to implement, very fast to compute even in streaming fashion, and easy to analyze. However, its density is almost a factor of 2 away from the lower bound for large windows. Methods. In this work we introduce mod-sampling, a two-step sampling algorithm to obtain new minimizer schemes. Given a (small) parameter t, the mod-sampling algorithm finds the position p of the smallest t-mer in a window. It then samples the k-mer at position pod w. The lr-minimizer uses t = k-w and the mod-minimizer uses t≡ k (mod w). Results. These new schemes have provably lower density than random minimizers and other schemes when k is large compared to w, while being as fast to compute. Importantly, the mod-minimizer achieves optimal density when k → ∞. Although the mod-minimizer is not the first method to achieve optimal density for large k, its proof of optimality is simpler than previous work. We provide pseudocode for a number of other methods and compare to them. In practice, the mod-minimizer has considerably lower density than the random minimizer and other state-of-the-art methods, like closed syncmers and miniception, when k > w. We plugged the mod-minimizer into SSHash, a k-mer dictionary based on minimizers. For default parameters (w,k) = (11,21), space usage decreases by 15% when indexing the whole human genome (GRCh38), while maintaining its fast query time.Source: LEIBNIZ INTERNATIONAL PROCEEDINGS IN INFORMATICS, vol. 312. Royal Holloway, London, United Kingdom, 2-4/09/2024
DOI: 10.4230/lipics.wabi.2024.11
Metrics:
See at:
drops.dagstuhl.de | CNR IRIS | CNR IRIS
2024
Book
Open Access
PerconAI 2024: 3rd Workshop on Pervasive and Resource-Constrained Artificial Intelligence - Welcome and Committees
Angelov P., Bernardi M. L., Nardini F. M., Pecori R., Valerio L.
PerconAI 2024 - Welcome and CommitteesDOI: 10.1109/percomworkshops59983.2024.10503466
Metrics:
Angelov P., Bernardi M. L., Nardini F. M., Pecori R., Valerio L.
PerconAI 2024 - Welcome and CommitteesDOI: 10.1109/percomworkshops59983.2024.10503466
Metrics:
See at:
CNR IRIS | ieeexplore.ieee.org | doi.org | CNR IRIS
2024
Journal article
Open Access
Where the patterns are: repetition-aware compression for colored de Bruijn graphs
Campanelli A., Pibiri G. E., Fan J., Patro R.
We describe lossless compressed data structures for the colored de Bruijn graph (or c-dBG). Given a collection of reference sequences, a c-dBG can be essentially regarded as a map from k-mers to their color sets. The color set of a k-mer is the set of all identifiers, or colors, of the references that contain the k-mer. While these maps find countless applications in computational biology (e.g., basic query, reading mapping, abundance estimation, etc.), their memory usage represents a serious challenge for large-scale sequence indexing. Our solutions leverage on the intrinsic repetitiveness of the color sets when indexing large collections of related genomes. Hence, the described algorithms factorize the color sets into patterns that repeat across the entire collection and represent these patterns once instead of redundantly replicating their representation as would happen if the sets were encoded as atomic lists of integers. Experimental results across a range of datasets and query workloads show that these representations substantially improve over the space effectiveness of the best previous solutions (sometimes, even dramatically, yielding indexes that are smaller by an order of magnitude). Despite the space reduction, these indexes only moderately impact the efficiency of the queries compared to the fastest indexes.Source: JOURNAL OF COMPUTATIONAL BIOLOGY
DOI: 10.1089/cmb.2024.0714
DOI: 10.1101/2024.07.09.602727
Project(s): EFRA
Metrics:
Campanelli A., Pibiri G. E., Fan J., Patro R.
We describe lossless compressed data structures for the colored de Bruijn graph (or c-dBG). Given a collection of reference sequences, a c-dBG can be essentially regarded as a map from k-mers to their color sets. The color set of a k-mer is the set of all identifiers, or colors, of the references that contain the k-mer. While these maps find countless applications in computational biology (e.g., basic query, reading mapping, abundance estimation, etc.), their memory usage represents a serious challenge for large-scale sequence indexing. Our solutions leverage on the intrinsic repetitiveness of the color sets when indexing large collections of related genomes. Hence, the described algorithms factorize the color sets into patterns that repeat across the entire collection and represent these patterns once instead of redundantly replicating their representation as would happen if the sets were encoded as atomic lists of integers. Experimental results across a range of datasets and query workloads show that these representations substantially improve over the space effectiveness of the best previous solutions (sometimes, even dramatically, yielding indexes that are smaller by an order of magnitude). Despite the space reduction, these indexes only moderately impact the efficiency of the queries compared to the fastest indexes.Source: JOURNAL OF COMPUTATIONAL BIOLOGY
DOI: 10.1089/cmb.2024.0714
DOI: 10.1101/2024.07.09.602727
Project(s): EFRA
Metrics:
See at:
Archivio istituzionale della ricerca - Università degli Studi di Venezia Ca' Foscari | Archivio istituzionale della ricerca - Università degli Studi di Venezia Ca' Foscari | IRIS Cnr | IRIS Cnr | Journal of Computational Biology | doi.org | CNR IRIS | CNR IRIS
2024
Conference article
Open Access
Optimizing resource allocation in the edge: a minimum weighted vertex cover approach
Makris A., Maragkoudakis E., Kontopoulos I., Theodoropoulos T., Korontanis I., Carlini E., Mordacchini M., Dazzi P., Varvarigou T.
The transition from Cloud Computing to a Cloud-Edge continuum introduces novel opportunities and challenges for data-intensive and interactive Next Generation applications. Strategies that were effective in the Cloud environment now necessitate reconsideration to adapt to the Edge's distributed, dynamic, and heterogeneous ecosystem. Proactively planning the placement of application images becomes crucial to minimize image transfer time, align with the dynamic nature of the system, and meet the strict demands of applications. In this regard, this paper proposes an empirical experimental analysis, by comparing the results of different placement strategies and various network topologies. In particular, we model the problem of proactive placement of application images as a Minimum Weighted Vertex Cover problem. Our results demonstrate that the Greedy approach seems to offer the optimal tradeoff with respect to the number of allocated images and execution time.DOI: 10.1145/3659994.3660316
Project(s): CHARITY
Metrics:
Makris A., Maragkoudakis E., Kontopoulos I., Theodoropoulos T., Korontanis I., Carlini E., Mordacchini M., Dazzi P., Varvarigou T.
The transition from Cloud Computing to a Cloud-Edge continuum introduces novel opportunities and challenges for data-intensive and interactive Next Generation applications. Strategies that were effective in the Cloud environment now necessitate reconsideration to adapt to the Edge's distributed, dynamic, and heterogeneous ecosystem. Proactively planning the placement of application images becomes crucial to minimize image transfer time, align with the dynamic nature of the system, and meet the strict demands of applications. In this regard, this paper proposes an empirical experimental analysis, by comparing the results of different placement strategies and various network topologies. In particular, we model the problem of proactive placement of application images as a Minimum Weighted Vertex Cover problem. Our results demonstrate that the Greedy approach seems to offer the optimal tradeoff with respect to the number of allocated images and execution time.DOI: 10.1145/3659994.3660316
Project(s): CHARITY
Metrics:
See at:
IRIS Cnr | doi.org | IRIS Cnr | Archivio della Ricerca - Università di Pisa | Archivio della Ricerca - Università di Pisa | CNR IRIS