A5022658803- Neural Networks and Applications
- Topic Modeling
- Advanced Image and Video Retrieval Techniques
- Machine Learning and Algorithms
- Natural Language Processing Techniques
- Web Data Mining and Analysis
- Advanced Graph Neural Networks
- Image Retrieval and Classification Techniques
- Fuzzy Logic and Control Systems
- Face and Expression Recognition
- Machine Learning and Data Classification
- Bayesian Modeling and Causal Inference
- Domain Adaptation and Few-Shot Learning
- Semantic Web and Ontologies
- Visual Attention and Saliency Detection
- Multimodal Machine Learning Applications
- Algorithms and Data Compression
- Explainable Artificial Intelligence (XAI)
- Constraint Satisfaction and Optimization
- SARS-CoV-2 and COVID-19 Research
- Advanced Vision and Imaging
- Model Reduction and Neural Networks
- Music and Audio Processing
- AI-based Problem Solving and Planning
- Graph Theory and Algorithms
University of Siena
2016-2025
Université Côte d'Azur
2020-2023
Centre National de la Recherche Scientifique
2020-2023
Institut national de recherche en informatique et en automatique
2020-2023
University of Padua
2023
Observatoire de la Côte d’Azur
2020-2022
Conference Board
2021-2022
Institut de Biologie Valrose
2022
Florence (Netherlands)
2019-2020
Universidade Federal do Rio Grande do Sul
2019
In several applications the information is naturally represented by graphs. Traditional approaches cope with graphical data structures using a preprocessing phase which transforms graphs into set of flat vectors. However, in this way, important topological may be lost and achieved results heavily depend on stage. This paper presents new neural model, called graph network (GNN), capable directly processing GNNs extends recursive networks can applied most practically useful kinds graphs,...
The authors propose a theoretical framework for backpropagation (BP) in order to identify some of its limitations as general learning procedure and the reasons success several experiments on pattern recognition. first important conclusion is that examples can be found which BP gets stuck local minima. A simple example get during gradient descent without having learned entire training set presented. This guarantees existence solution with null cost. Some conditions network architecture...
Although the interest of a Web page is strictly related to its content and subjective readers' cultural background, measure authority can be provided that only depends on topological structure Web. PageRank noticeable way attach score pages basis connectivity. In this article, we look inside disclose fundamental properties concerning stability, complexity computational scheme, critical role parameters involved in computation. Moreover, introduce circuit analysis allows us understand...
A structured organization of information is typically required by symbolic processing. On the other hand, most connectionist models assume that data are organized according to relatively poor structures, like arrays or sequences. The framework described in this paper an attempt unify adaptive artificial neural nets and belief for problem processing information. In particular, relations between variables expressed directed acyclic graphs, where both numerical categorical values coexist....
Learning from examples plays a central role in artificial neural networks. The success of many learning schemes is not guaranteed, however, since algorithms like backpropagation may get stuck local minima, thus providing suboptimal solutions. For feedforward networks, optimal can be achieved provided that certain conditions on the network and environment are met. This principle investigated for case networks using radial basis functions (RBF). It assumed patterns separable by hyperspheres....
In this paper, we investigate the capabilities of local feedback multilayered networks, a particular class recurrent in which connections are only allowed from neurons to themselves. class, learning can be accomplished by an algorithm that is both space and time. We describe limits properties these networks give some insights on their use for solving practical problems.
Artificial neural networks have been extensively applied to document analysis and recognition. Most efforts devoted the recognition of isolated handwritten printed characters with widely recognized successful results. However, many other processing tasks, like preprocessing, layout analysis, character segmentation, word recognition, signature verification, effectively faced very promising This paper surveys most significant problems in area offline image processing, where connectionist-based...
Every day researchers from all over the world have to filter huge mass of existing research papers with crucial aim finding out useful publications related their current work. In this paper we propose a recommending algorithm based on citation graph and random-walker properties. The PaperRank is able assign preference score set documents contained in digital library linked one each other by bibliographic references. A data extracted ACM portal has been used for testing very promising...
Current advances in Artificial Intelligence and machine learning general, deep particular have reached unprecedented impact not only across research communities, but also over popular media channels. However, concerns about interpretability accountability of AI been raised by influential thinkers. In spite the recent AI, several works identified need for principled knowledge representation reasoning mechanisms integrated with learning-based systems to provide sound explainable models such...
Neural-symbolic computing has now become the subject of interest both academic and industry research laboratories. Graph Neural Networks (GNNs) have been widely used in relational symbolic domains, with widespread application GNNs combinatorial optimization, constraint satisfaction, reasoning other scientific domains. The need for improved explainability, interpretability trust AI systems general demands principled methodologies, as suggested by neural-symbolic computing. In this paper, we...
Deep learning allows to develop feature representations and train classification models in a fully integrated way. However, deep networks is quite hard it improves over shallow architectures only if large number of training data available. Injecting prior knowledge into the learner principled way reduce amount required data, as does not need induce from itself. In this paper we propose general integrate when networks. Semantic Based Regularization (SBR) used underlying framework represent...
Many pathogens exploit host cell-surface glycans. However, precise analyses of glycan ligands binding with heavily modified pathogen proteins can be confounded by overlapping sugar signals and/or compounded known experimental constraints. Universal saturation transfer analysis (uSTA) builds on existing nuclear magnetic resonance spectroscopy to provide an automated workflow for quantitating protein-ligand interactions. uSTA reveals that early-pandemic, B-origin-lineage severe acute...
Ensuring software reliability through early-stage defect prevention and prediction is crucial, particularly as systems become increasingly complex. Automated testing has emerged the most practical approach to achieving bug-free efficient code. In this context, machine learning-driven methods, especially those leveraging natural language models, have gained significant traction for developing effective techniques. This paper introduces a novel framework automating prediction, focusing on...
In this paper, we propose a general framework for graph matching which is suitable different problems of pattern recognition. The representation assume at the same time highly structured, like classic syntactic and structural approaches, subsymbolic nature with real-valued features, connectionist statistic approaches. We show that random walk based models, inspired by Google's PageRank, give rise to spectral theory nicely enhances topological features node level. As straightforward...
This paper describes the neural-based recognition and verification techniques used in a banknote machine, recently implemented for accepting currency of different countries. The perception mechanism is based on low-cost optoelectronic devices which produce signal associated with light refracted by banknotes. classification steps are carried out society multilayer perceptrons whose operation properly scheduled an external controlling algorithm, guarantees real-time implementation standard...
Within the GEN-COVID Multicenter Study, biospecimens from more than 1000 SARS-CoV-2 positive individuals have thus far been collected in Biobank (GCB). Sample types include whole blood, plasma, serum, leukocytes, and DNA. The GCB links samples to detailed clinical data available Patient Registry (GCPR). It includes hospitalized patients (74.25%), broken down into intubated, treated by CPAP-biPAP, with O
Explainable artificial intelligence has rapidly emerged since lawmakers have started requiring interpretable models for safety-critical domains. Concept-based neural networks arisen as explainable-by-design methods they leverage human-understandable symbols (i.e. concepts) to predict class memberships. However, most of these approaches focus on the identification relevant concepts but do not provide concise, formal explanations how such are leveraged by classifier make predictions. In this...
Classification is an important problem in image document processing and often a preliminary step toward recognition, understanding, information extraction. In this paper, the formulated framework of concept learning each category corresponds to set documents with similar physical structure. We propose solution based on two algorithmic ideas. First, we obtain structured representation images labeled XY-trees (this informs learner about relationships between subconstituents). Second,...