As the cognitive engine of urban Cyber-Physical Systems (CPS), accurate traffic prediction is essential for optimizing physical flows through cybernetic control. Cross-city traffic prediction is critical for intelligent transportation systems but faces a fundamental dilemma: purely local training suffers from data scarcity in emerging cities, while centralized training compromises data privacy. FL offers a decentralized learning alternative, yet traditionally struggles with significant communication overhead and domain heterogeneity. Existing methods typically rely on iterative, multi-round parameter synchronization to align disparate traffic patterns, which is often infeasible in bandwidth-constrained cyber–physical systems. In this paper, we propose SWIFT ( S patiotemporal W eighed I ntegration for F ederated T raffic Prediction), a communication-efficient framework that achieves high-performance traffic prediction with only One-shot of communication. Unlike methods that blindly aggregate these conflicting features, SWIFT explicitly decomposes them into orthogonal semantic subspaces. This disentanglement ensures that only semantically aligned temporal knowledge is shared, effectively eliminating the need for iterative correction. Furthermore, the framework incorporates a context-aware dynamic gating mechanism that adaptively weights and fuses these decoupled features according to real-time traffic conditions. Extensive experiments on four real-world datasets demonstrate that SWIFT matches the performance of mature multi-round FL methods, offering a scalable paradigm for data-scarce urban environments.

Knowledge Graphs (KGs) offer an effective framework for organizing complex information, yet their construction and use for question answering face significant challenges due to the dynamic and evolving nature of research domains. In this work, we present MAKES-QA , a modular and dynamic multi-agent framework for KG-based question answering (KGQA) that leverages Large Language Models (LLMs) to extract, normalize, and integrate knowledge, particularly in the context of scientific literature. The framework is explicitly designed as a research support tool for domain-aware users, such as researchers and practitioners, enabling guided and informed interaction during knowledge graph construction and exploration. The system enables iterative enrichment of the KG through user-guided strategies, facilitating the discovery and integration of new information. Once constructed, the KG supports natural language queries via a Retrieval-Augmented Generation (RAG) approach that combines semantic retrieval of relevant triples with LLM-based answer synthesis. This architecture ensures accurate, context-aware responses grounded in curated scientific knowledge. Our approach promotes efficient exploration of scientific domains, reducing the need for exhaustive manual reading while enabling flexible knowledge discovery. The MAKES-QA source code is available at https://github.com/MODAL-UNINA/MAKES-QA .

One-Shot Federated Learning (OSFL) enables collaborative training in a single round, eliminating the need for iterative communication, making it particularly suitable for use in resource-constrained and privacy-sensitive applications. This survey offers a thorough examination of One-Shot FL, highlighting its distinct operational framework compared to traditional federated approaches. One-Shot FL supports resource-limited devices by enabling single-round model aggregation while maintaining data locality. The survey systematically categorizes existing methodologies, emphasizing advancements in client model initialization, aggregation techniques, and strategies for managing heterogeneous data distributions. Furthermore, we analyze the limitations of current approaches, particularly in terms of scalability and generalization in non-IID settings. By analyzing cutting-edge techniques and outlining open challenges, this survey aims to provide a comprehensive reference for researchers and practitioners seeking to design and implement One-Shot FL systems, advancing the development and adoption of One-Shot FL solutions in real-world, resource-constrained settings.

Federated learning (FL) enables the collaborative training of artificial intelligence models across multiple participating clients while preserving data privacy. Yet, the presence of statistical heterogeneity, characterized by non-independent and non-identically distributed (non-IID) data among clients, poses a significant hurdle in achieving optimal model convergence within the federated setting. In this study, we present FLOWS, a framework that seamlessly incorporates the Sinkhorn distance into each client’s local training process. This integration effectively tackles the well-known challenge by promoting a close alignment between local predictions and the global model’s predictions. Comprehensive experiments across diverse datasets were conducted to evaluate FLOWS’s performance against state-of-the-art FL algorithms. The results indicate that FLOWS enhances the performance of FL models without incurring in a substantial additional computational load.

Abstract Nowadays, the great diffusion of the Internet of Things and the improvements in Artificial Intelligence techniques have given a rise in the development and application of data‐driven approaches for Predictive Maintenance to reduce the costs linked to the maintenance of industrial machinery. Due to the wide real‐life applications and the strong interest by even more industries, this field is highly attractive for academics and practitioners. So, constructing efficient frameworks to address the Predictive Maintenance problem is an open debate. In this work, we propose a Deep Learning approach for the feature extraction in the offshore oil wells monitoring context, exploiting the public 3 W dataset, which is well‐known in the literature. The dataset is made up of about 2000 multivariate time series labelled according to the corresponding functioning of the well. So, there is a classification task with eight classes, each related to a particular machinery condition. Thanks to the peculiarities of the labels, the proposed framework is valid both for diagnostics and prognostics. In more detail, we compare two different approaches in feature extraction. The first is a statistical approach, widely used in the literature related to the considered dataset; the second is based on Convolutional 1D AutoEncoder. The extracted features are then used as input for several Machine Learning algorithms, namely the Random Forest, Nearest Neighbours, Gaussian Naive Bayes and Quadratic Discriminant Analysis. Different experiments on various time horizons prove the worthiness of the Convolutional AutoEncoder.

Network Intrusion Detection Systems (NIDS) are crucial tools for protecting networked devices from cyberattacks. Recent development in the field of Artificial Intelligence (AI) has provided tremendous advantages in implementing NIDSs able to monitor network traffic and block cyberattacks in real-time. In the literature, it is widely recognized that the effective training of a NIDS requires a large quantity of labeled traffic, representative of attacks. Nonetheless, the availability of public and abundant datasets remains remarkably restricted due to the cost of gathering and labeling real traffic traces and privacy concerns for sharing them. To tackle these challenges, in this paper we present a generative AI model capable of synthesizing anonymized traffic traces from real ones, thus dealing with privacy, abundance, and representativeness. The proposal is based on a Conditional Variational Autoencoder (CVAE) and a preprocessing procedure specifically designed for the generation of new traffic traces. To validate our solution, we conduct an extensive empirical study leveraging three recent and publicly-available datasets, containing benign and malicious traffic. The validation is carried out from both the perspectives of classification performance of a robust NIDS and the quality of synthetic data, in comparison to the utilization of real data. We compare our CVAE with two state-of-the-art AI-based traffic data generators and prove that, trained with traces emitted by our generative model, a NIDS has a limited F1-score loss compared to training on real data; competing models instead struggle or fail to generate traces that are as effective for NIDS training and as statistically similar to the original. We make the synthetic datasets available in both PCAP and tabular formats, to facilitate the reproducibility of our findings and encourage further exploration in the field of generative AI for networking.

Environment perception through deep computation in unstructured environments is important for the construction of autonomous navigation systems. Most research focuses on navigation in structured scenes, including indoor mobility and driving along roads, while neglecting to consider unstructured environments, which often contain diverse heights and distributions. In addition, existing depth estimation algorithms based on deep learning often need to complete training under the supervision of Ground truth, and GT data with a large number of labels are not always easy to obtain. To tackle this issue, this paper proposes an unsupervised stereo depth estimation method for processing UAV navigation images in an unstructured environment. The method contains a primitive U-shaped CNN network architecture for processing such scenes. The feature extraction layer of the network is based on the YOLOv3 residual structure, and additional attention modules help the network enhance its ability to perceive image features. Finally, depth estimation experiments on the unstructured environments dataset Mid-Air further demonstrate the effectiveness and reliability of the proposed method.

Anonymization is an important aspect of data privacy protection, especially in the context of sensitive personal information collected through sensors. In this paper, we propose a new service-based architecture for anonymizing such data in real-time, ensuring that data is accessible to authorized users while maintaining privacy. Our architecture is based on the annotation of data at ingestion time, where privacy levels are assigned to sets of columns. The anonymization procedure is performed by compressing and encoding the data through an autoencoder model, where the encoder and decoder functions are defined as parametric functions composed of multiple hidden layers.

The Forward-forward (FF) algorithm is a new method for training neural networks, proposed as an alternative to the traditional Backpropagation (BP) algorithm by Hinton. The FF algorithm replaces the backward computations in the learning process with another forward pass. Each layer has an objective function, which aims to be high for positive data and low for negative ones. This paper presents a preliminary investigation into variations of the FF algorithm, such as incorporating a local Backpropagation to create a hybrid network that robustly converges while preserving the ability to avoid backward computations when needed, for example, in non-differentiable areas of the network. Additionally, a pseudo-random logic for selecting trainable stacks of layers at each epoch is proposed to speed up the learning process.

Characterizing the heterogeneity of cancer metabolism requires the knowledge of metabolic fluxes in different tumor types. These fluxes cannot be directly determined, especially at a sub-cellular level. Still, they can be obtained numerically through constraint-based steady-state models after integrating other high-throughput -omics data, such as transcriptomics. In this work, we proposed to study cancer metabolism through data analysis and machine learning methodologies. To this aim, we considered transcriptomics profiles for a large set of cancer cells. Using a core metabolic network as a scaffold, we generated many feasible flux distributions for each cancer cell. Then, we used cluster analysis to analyze these data. This preliminary analysis revealed three well-separated clusters having different metabolic behaviors.

Over the past two years, the COVID-19 pandemic has been one of the most frequently and hotly debated social topics. Lockdowns and restrictions radically change the way of working and socializing due to social distancing and wearing masks; the ongoing pandemic impacts people's life and psychological health. Infection Rate Rthas been the main parameter used by national and local governments worldwide for describing the pandemic behavior synthetically. Rtwas adopted to define containment policies (lockdowns, social distancing, intermittent regional strategies, etc.) that have affected social life. In the present paper, we propose an Artificial Intelligence (AI) approach for the modeling of the COVID-19 Infection Rate Rtby exploiting the novel methodology of the Physics-Informed Neural Networks (PINNs) to compute the susceptible-infected-dead-recovered (SIDR) model. To test the accuracy of the neural network, we predicted the susceptible, infected, dead, and recovered on the next 30 days against the considered period.

The advent of the digital innovation era is changing service, use, and resources management paradigms, offering a wide range of new and essential opportunities. In particular, the advent of the Internet of Things (IoT), i.e., the ability to connect individual objects to the Internet, also capable of communicating autonomously, has its particular declination on the connected vehicle. It is combined with the potential of advanced sensors placed pervasively on vehicles, which offer multifunctional monitoring capabilities of the entire system: from individual components up to the whole vehicle, including driver behavior and conditions and many exogenous parameters to the vehicle (road and weather conditions, congestion, risk situations, changes to mobility plans, etc.). In this perspective, machine learning (ML) models can transform raw data into new knowledge; they can contribute in an innovative way to define and suggest decisions, strategies, and criteria for resource use. Nowadays, most intelligent mobility projects also integrate artificial intelligence (AI) and ML solutions. In this article, we present and discuss the application of unsupervised learning techniques on a vehicular IoT data set. The main goal is to generate new knowledge about a geographical zone by analyzing historical drivers behavioral data. The autonomous vehicle’s framework can exploit the generated valuable insights to optimize the routes and prevent critical issues.

Abstract Nowadays, statistical arbitrage is one of the most attractive fields of study for researchers, and its applications are widely used also in the financial industry. In this work, we propose a new approach for statistical arbitrage based on clustering stocks according to their exposition on common risk factors. A linear multifactor model is exploited as theoretical background. The risk factors of such a model are extracted via Principal Component Analysis by looking at different time granularity. Furthermore, they are standardized to be handled by a feature selection technique, namely the Adaptive Lasso, whose aim is to find the factors that strongly drive each stock’s return. The assets are then clustered by using the information provided by the feature selection, and their exposition on each factor is deleted to obtain the statistical arbitrage. Finally, the Sequential Least SQuares Programming is used to determine the optimal weights to construct the portfolio. The proposed methodology is tested on the Italian, German, American, Japanese, Brazilian, and Indian Stock Markets. Its performances, evaluated through a Cross-Validation approach, are compared with three benchmarks to assess the robustness of our strategy.

Deep learning technology has become one of the core driving forces to promote the in-depth development of industrial automation. In [A1], Wang et al. interpreted the decision process of the convolutional neural network (CNN) by constructing a percolation model from a statistical physics perspective. In this perspective, the decision-making basis of CNN is difficult to understand, because CNN is usually used as a black box model. Furthermore, a novel concept of the differentiation degree and summarized an empirical formula for quantifying the differentiation degree is presented and discussed.

New technologies are transforming medicine, and this revolution starts with data. Usually, health services within public healthcare systems are accessed through a booking centre managed by local health authorities and controlled by the regional government. In this perspective, structuring e-health data through a Knowledge Graph (KG) approach can provide a feasible method to quickly and simply organize data and/or retrieve new information. Starting from raw health bookings data from the public healthcare system in Italy, a KG method is presented to support e-health services through the extraction of medical knowledge and novel insights. By exploiting graph embedding which arranges the various attributes of the entities into the same vector space, we are able to apply Machine Learning (ML) techniques to the embedded vectors. The findings suggest that KGs could be used to assess patients' medical booking patterns, either from unsupervised or supervised ML. In particular, the former can determine possible presence of hidden groups of entities that is not immediately available through the original legacy dataset structure. The latter, although the performance of the used algorithms is not very high, shows encouraging results in predicting a patient's likelihood to undergo a particular medical visit within a year. However, many technological advances remain to be made, especially in graph database technologies and graph embedding algorithms.

Abstract Compared to other features of the human body, voice is quite complex and dynamic, in a sense that a speech can be spoken in various languages with different accents and in different emotional states. Recognizing the gender, i.e. male or female from the voice of an individual, is by all accounts a minor errand for human beings. Similar goes for speaker identification if we are well accustomed with the speaker for a long time. Our ears function as the front end, accepting the sound signs which our cerebrum processes and settles on our disposition. Although being trivial for us, it becomes a challenging task to mimic for any computing device. Automatic gender, emotion and speaker identification systems have many applications in surveillance, multimedia technology, robotics and social media. In this paper, we propose a Golden Ratio-aided Neural Network (GRaNN) architecture for the said purposes. As deciding the number of units for each layer in deep NN is a challenging issue, we have done this using the concept of Golden Ratio. Prior to that, an optimal subset of features are selected from the feature vector extracted, common for all three tasks, from spectral images obtained from the input voice signals. We have used a wrapper-filter framework where minimum redundancy maximum relevance selected features are fed to Mayfly algorithm combined with adaptive beta hill climbing (A $$\beta$$ β HC) algorithm. Our model achieves accuracies of 99.306% and 95.68% for gender identification in RAVDESS and Voice Gender datasets, 95.27% for emotion identification in RAVDESS dataset and 67.172% for speaker identification in RAVDESS dataset. Performance comparison of this model with existing models on the publicly available datasets confirms its superiority over those models. Results also ensure that we have chosen the common feature set meticulously, which works equally well on three different pattern classification tasks. The proposed wrapper-filter framework reduces the feature dimension significantly, thereby lessening the storage requirement and training time. Finally, strategically selecting the number units in each layer in NN help increases the overall performance of all three pattern classification tasks.

Nowadays, predictive medicine begins to become a reality thanks to Artificial Intelligence (AI) which allows, through the processing of huge amounts of data, to identify correlations not perceptible to the human brain. The application of AI in predictive diagnostics is increasingly pervasive; through the use and interpretation of data, the first signs of some diseases (i.e. tumours) can be detected to help physicians make more accurate diagnoses to reduce the errors and develop methods for individualized medical treatment. In this perspective, salivary gland tumours (SGTs) are rare cancers with variable malignancy representing less than 1% of all cancer diagnoses and about 5% of head and neck cancers. The clinical management of SGTs is complicated by a high rate of preclinical diagnostic errors. Today, fine needle aspiration cytology (FNAC) represents the primary diagnostic tool in the hands of clinicians. However, it provides information that about 25% of cases are dubious or inconclusive, complicating therapeutic choices. Thus, finding new tools supporting clinicians to make the right choices in doubtful cases is necessary. This research work presents and discusses a Deep Learning-based framework for automatic segmentation and classification of salivary gland tumours. Furthermore, we propose an explainable segmentation learning approach supporting the effectiveness of the proposed framework through a per-epoch learning process analysis and the attention map mechanism. The proposed framework was evaluated with a collected CT dataset of patients with salivary gland tumours. Experimental results show that our methodology achieves significant scores on both segmentation and classification tasks.

The outbreak of a global pandemic called coronavirus has created unprecedented circumstances resulting into a large number of deaths and risk of community spreading throughout the world. Desperate times have called for desperate measures to detect the disease at an early stage via various medically proven methods like chest computed tomography (CT) scan, chest X-Ray, etc., in order to prevent the virus from spreading across the community. Developing deep learning models for analysing these kinds of radiological images is a well-known methodology in the domain of computer based medical image analysis. However, doing the same by mimicking the biological models and leveraging the newly developed neuromorphic computing chips might be more economical. These chips have been shown to be more powerful and are more efficient than conventional central and graphics processing units. Additionally, these chips facilitate the implementation of spiking neural networks (SNNs) in real-world scenarios. To this end, in this work, we have tried to simulate the SNNs using various deep learning libraries. We have applied them for the classification of chest CT scan images into COVID and non-COVID classes. Our approach has achieved very high F1 score of 0.99 for the potential-based model and outperforms many state-of-the-art models. The working code associated with our present work can be found here.

Nowadays, a sustainable and smart city focuses on energy efficiency and the reduction of polluting emissions through smart mobility projects and initiatives to “sensitize” infrastructure. Smart parking is one of the building blocks of intelligent mobility, innovative mobility that aims to be flexible, integrated, and sustainable and consequently integrated into a Smart City. By using the Internet of Things (IoT) sensors located in the parking areas or the underground car parks in combination with a mobile application, which indicates to citizens the free places in the different areas of the city and guides them toward the chosen parking, it is possible to reduce air pollution and fluidifying noise traffic. In this article, we present and discuss an innovative Deep Learning-based ensemble technique in forecasting the parking space occupancy to reduce the search time for parking and to optimize the flow of cars in particularly congested areas, with an overall positive impact on traffic in urban centres. A genetic algorithm has also been used to optimize predictors parameters. The main goal is to design an intelligent IoT-based service that can predict, in the next few hours, the parking spaces occupancy of a street. The proposed approach has been assessed on a real IoT dataset composed by over than 15M of collected sensor records. Obtained results demonstrate that our method outperforms both single predictors and the widely used strategy of the mean providing inherently robust predictions.