A5066187890- Machine Learning and Algorithms
- Machine Learning and Data Classification
- Bayesian Modeling and Causal Inference
- Explainable Artificial Intelligence (XAI)
- Machine Learning in Bioinformatics
- Data Management and Algorithms
- Protein Structure and Dynamics
- Semantic Web and Ontologies
- Natural Language Processing Techniques
- Topic Modeling
- AI-based Problem Solving and Planning
- Neural Networks and Applications
- Constraint Satisfaction and Optimization
- RNA and protein synthesis mechanisms
- Data Stream Mining Techniques
- Multi-Criteria Decision Making
- Rough Sets and Fuzzy Logic
- Adversarial Robustness in Machine Learning
- Computational Drug Discovery Methods
- Domain Adaptation and Few-Shot Learning
- Mobile Crowdsensing and Crowdsourcing
- Software Engineering Research
- Complex Network Analysis Techniques
- Advanced Multi-Objective Optimization Algorithms
- Multimodal Machine Learning Applications
University of Trento
2016-2025
ServiceNow (United States)
2024
Delft University of Technology
2024
University of Bologna
2024
Aalborg University
2021
Laboratoire d'Informatique de Paris-Nord
2021
DePaul University
2015
University of Potsdam
2009
University of Florence
2001-2008
KU Leuven
2008
Deep learning (DL) has proved successful in medical imaging and, the wake of recent COVID-19 pandemic, some works have started to investigate DL-based solutions for assisted diagnosis lung diseases. While existing focus on CT scans, this paper studies application DL techniques analysis ultrasonography (LUS) images. Specifically, we present a novel fully-annotated dataset LUS images collected from several Italian hospitals, with labels indicating degree disease severity at frame-level,...
DISULFIND is a server for predicting the disulfide bonding state of cysteines and their connectivity starting from sequence alone. Optionally, can be predicted assignment given as input. The output simple visualization assigned (with confidence degrees) most likely patterns. available at http://disulfind.dsi.unifi.it/ .
Abstract Background The classical view on eukaryotic gene expression proposes the scheme of a forward flow for which fluctuations in mRNA levels upon stimulus contribute to determine variations availability translation. Here we address this issue by simultaneously profiling with microarrays total mRNAs (the transcriptome) and polysome-associated translatome) after EGF treatment human cells, extending analysis other 19 different transcriptome/translatome comparisons mammalian cells following...
Normative texts can be viewed as composed by formal partitions (articles, paragraphs, etc.) or semantic units containing fragments of a regulation (provisions). Provisions described according to metadata scheme which consists provision types and their arguments. This annotation normative text make the retrieval norms easier. The detection description provisions established is an analytic intellectual activity aiming at classifying portions into extract Automatic facilities supporting this...
We study the problem of multiclass classification within framework error correcting output codes (ECOC) using margin-based binary classifiers. Specifically, we address two important open problems in this context: decoding and model selection. The concerns how to map outputs classifiers into class codewords. In paper introduce a new function that combines margins through an estimate their conditional probabilities. Concerning selection, present theoretical results bounding leave-one-out (LOO)...
The centrality of the decision maker (DM) is widely recognized in multiple criteria decision-making community. This translates into emphasis on seamless human-computer interaction, and adaptation solution technique to knowledge which progressively acquired from DM. paper adopts methodology reactive search optimization (RSO) for evolutionary interactive multiobjective optimization. RSO follows paradigm "learning while optimizing," through use online machine learning techniques as an integral...
Metalloproteins are proteins capable of binding one or more metal ions, which may be required for their biological function, regulation activities structural purposes. Metal-binding properties remain difficult to predict as well investigate experimentally at the whole-proteome level. Consequently, current knowledge about metalloproteins is only partial.The present work reports on development a machine learning method prediction zinc-binding state pairs nearby amino-acids, using predictors...
The rapid dynamics of COVID-19 calls for quick and effective tracking virus transmission chains early detection outbreaks, especially in the "phase 2" pandemic, when lockdown other restriction measures are progressively withdrawn, order to avoid or minimize contagion resurgence. For this purpose, contact-tracing apps being proposed large scale adoption by many countries. A centralized approach, where data sensed app all sent a nation-wide server, raises concerns about citizens' privacy...
Abstract Temporal networks are essential for modeling and understanding time-dependent systems, from social interactions to biological systems. However, real-world data construct meaningful temporal expensive collect or unshareable due privacy concerns. Generating arbitrarily large anonymized synthetic graphs with the properties of networks, namely surrogate is a potential way bypass problem. it not easy build which do lack information on and/or topological input network their correlations....
Abstract Accurate predictions of metal‐binding sites in proteins by using sequence as the only source information can significantly help prediction protein structure and function, genome annotation, experimental determination structure. Here, we introduce a method for identifying histidines cysteines that participate binding several transition metals iron complexes. The predicts being either two states (free or metal bound) three (free, bound, disulfide bridges). uses utilizing...
MetalDetector identifies CYS and HIS involved in transition metal protein binding sites, starting from sequence alone. A major new feature of release 2.0 is the ability to predict which residues are jointly coordination same ion. The server available at http://metaldetector.dsi.unifi.it/v2.0/ .
While constraints are ubiquitous in artificial intelligence and also commonly used machine learning data mining, the problem of from examples has received less attention. In this paper, we discuss constraint detail, indicate some subtle differences with standard problems, sketch applications summarize state-of-the-art.
Graph Neural Networks (GNNs) have become the leading paradigm for learning on (static) graph-structured data. However, many real-world systems are dynamic in nature, since graph and node/edge attributes change over time. In recent years, GNN-based models temporal graphs emerged as a promising area of research to extend capabilities GNNs. this work, we provide first comprehensive overview current state-of-the-art GNN, introducing rigorous formalization settings tasks novel taxonomy...
The impressive performance of modern Large Language Models (LLMs) across a wide range tasks, along with their often non-trivial errors, has garnered unprecedented attention regarding the potential AI and its impact on everyday life. While considerable effort been continues to be dedicated overcoming limitations current models, potentials risks human-LLM collaboration remain largely underexplored. In this perspective, we argue that enhancing focus interaction should primary target for future...
Self-Explainable Graph Neural Networks (SE-GNNs) are popular explainable-by-design GNNs, but the properties and limitations of their explanations not well understood. Our first contribution fills this gap by formalizing extracted SE-GNNs, referred to as Trivial Explanations (TEs), comparing them established notions explanations, namely Prime Implicant (PI) faithful explanations. analysis reveals that TEs match PI for a restricted significant family tasks. In general, however, they can be...
Graph transformers extend global self-attention to graph-structured data, achieving notable success in graph learning. Recently, random walk structural encoding (RWSE) has been found further enhance their predictive power by both and positional information into the edge representation. However, RWSE cannot always distinguish between edges that belong different local patterns, which reduces its ability capture full complexity of graphs. This work introduces Simple Path Structural Encoding...
Concept-based Models are neural networks that learn a concept extractor to map inputs high-level concepts and an inference layer translate these into predictions. Ensuring modules produce interpretable behave reliably in out-of-distribution is crucial, yet the conditions for achieving this remain unclear. We study problem by establishing novel connection between reasoning shortcuts (RSs), common issue where models achieve high accuracy learning low-quality concepts, even when fixed provided...