A5009738158- Recommender Systems and Techniques
- Machine Learning and Data Classification
- Software System Performance and Reliability
- Privacy-Preserving Technologies in Data
- Business Process Modeling and Analysis
- Neural Networks and Applications
- Advanced Bandit Algorithms Research
- Advanced Graph Neural Networks
- Service-Oriented Architecture and Web Services
- Text and Document Classification Technologies
- Topic Modeling
- Stochastic Gradient Optimization Techniques
- Data Mining Algorithms and Applications
- Anomaly Detection Techniques and Applications
- Data Stream Mining Techniques
- Rough Sets and Fuzzy Logic
- Bayesian Modeling and Causal Inference
- Advanced Malware Detection Techniques
- Cryptography and Data Security
- Machine Learning and Algorithms
- Explainable Artificial Intelligence (XAI)
- Face and Expression Recognition
- Ferroelectric and Negative Capacitance Devices
- Human Pose and Action Recognition
- Human Mobility and Location-Based Analysis
University of Turin
2021-2025
University of Padua
2014-2022
Predicting the completion time of business process instances would be a very helpful aid when managing processes under service level agreement constraints. The ability to know in advance trend running allow managers react time, order prevent delays or undesirable situations. However, making such accurate forecasts is not easy: many factors may influence required complete instance. In this paper, we propose an approach based on deep Recurrent Neural Networks (specifically LSTMs) that able...
Accurate prediction of the completion time a business process instance would constitute valuable tool when managing processes under service level agreement constraints. Such prediction, however, is very challenging task. A wide variety factors could influence trend instance, and hence just using statistics historical cases cannot be sufficient to get accurate predictions. Here we propose new approach where, in order improve quality, both control data flow perspectives are jointly used. To...
E-commerce and online services are getting more ubiquitous day by day. Like many other e-commerce paradigms, grocery can highly benefit from recommender systems, especially when it comes to predicting users' shopping behavior. This specific scenario owns peculiar characteristics, such as repetitiveness loyalty, which makes the task very different standard recommendations. In this work, we present an efficient solution compute next basket recommendation, under a general top-n recommendation...
The latest developments in the field of artificial intelligence (AI) have given rise to many ethical and socio-economic concerns. Nonetheless, impact AI technologies is evident tangible our everyday life. This dichotomy leads mixed feelings toward AI: people recognize positive AI, but they also show concerns, especially about their privacy security. In this article, we try understand whether implicit explicit attitudes are coherent. We investigated by combining a self-report measure an...
Predictive current control schemes strongly rely on the knowledge of plant model. The accuracy prediction could be affected by parameters variation or mismatch, non idealities and other model inadequacies. In Synchronous Reluctance Machine this effect particularly critical since its inherent intense iron saturation causes indunctances in a wide range. For reason, nominal unsuitable for lead to severe deterioration drive performance. paper novel Model-Free Current Control is presented....
Federated Learning has been proposed to develop better AI systems without compromising the privacy of final users and legitimate interests private companies. Initially deployed by Google predict text input on mobile devices, FL in many other industries. Since its introduction, mainly exploited inner working neural networks gradient descent-based algorithms either exchanging weights model or gradients computed during learning. While this approach very successful, it rules out applying...
Higher-order interactions (HOIs) are ubiquitous in real-world complex systems and applications. Investigation of deep learning for HOIs, thus, has become a valuable agenda the data mining machine communities. As networks HOIs expressed mathematically as hypergraphs, hypergraph neural (HNNs) have emerged powerful tool representation on hypergraphs. Given emerging trend, we present first survey dedicated to HNNs, with an in-depth step-by-step guide. Broadly, overviews HNN architectures,...
With the surge in popularity of cryptocurrencies, Bitcoin has emerged as one most promising means for remittance, payments, and trading. Supplemented by convenience offered smartphones, an increasing number users are adopting wallet apps different purposes.
Research over how suicide survivors approach services is limited. Aims: This cross-sectional study explores the psychological state and perceived social support of Italian survivors, including those who have not sought for help, investigates differences gender or kinship with departed. Methods: Rule-based system (RBS) analyses identified relationships between reported formal/informal help-seeking behavior. One-hundred thirty-two (103F; 27M) (53 having never support) answered an anonymous...
Decentralised Machine Learning (DML) enables collaborative machine learning without centralised input data. Federated (FL) and Edge Inference are examples of DML. While tools for DML (especially FL) starting to flourish, many not flexible portable enough experiment with novel processors (e.g., RISC-V), non-fully connected network topologies, asynchronous collaboration schemes. We overcome these limitations via a domain-specific language allowing us map schemes an underlying middleware, i.e....
Many natural substances and drugs are radical scavengers that prevent the oxidative damage to fundamental cell components. This process may occur via different mechanisms, among which, one of most important, is hydrogen atom transfer. The feasibility this can be assessed in silico using quantum mechanics compute ΔGHAT○. approach accurate, but time consuming. use machine learning (ML) allows us reduce tremendously computational cost assessment scavenging properties a potential antioxidant,...
In this paper we present our method used in the RecSys '16 Challenge.
Recommender Systems (RSs) are valuable technologies that help users in their decision-making process. Generally, RSs designed with the assumption a central server stores and manages historical users' behaviors. However, nowadays more aware of privacy issues leading to higher demand for privacy-preserving technologies. To cope this issue, Federated Learning (FL) paradigm can provide good performance without harming privacy. Some efforts have been devoted adapt standard collaborative filtering...
Deep generative modeling (DGM) is an increasingly popular approach that can create novel and unseen data, starting from a given data set. As the technology shows promising applications, many ethical issues also arise. For example, their misuse enable disinformation campaigns powerful phishing attempts. Research different biases affect deep learning models, leading to social such as misrepresentation. In this work, we formulate setting deal with similar problems, showing repurposed anomaly...
Kernel based classifiers, such as SVM, are considered state-of-the-art algorithms and widely used on many classification tasks. However, this kind of methods hardly interpretable for reason they often black-box models. In paper, we propose a new family Boolean kernels categorical data where features correspond to propositional formulas applied the input variables. The idea is create human-readable ease extraction interpretation rules directly from embedding space. Experiments artificial...