A carregar...
Projeto de investigação
LASIGE - Extreme Computing
Financiador
Autores
Publicações
A resilient continuous-time consensus method using a switching topology
Publication . Ramos, Guilherme; Silvestre, Daniel; Aguiar, A. Pedro; DEE - Departamento de Engenharia Electrotécnica e de Computadores; Elsevier
This paper addresses the design problem of a resilient consensus algorithm for agents with continuous-time dynamics. The main proposal is that by incorporating a switching mechanism selecting the network topology to avoid malicious nodes from communicating, the remaining nodes will converge to a value closer to the original steady-state without the attacker being present. The switching occurs at discrete-time steps where each node evaluates the reputation score of the neighbors and deactivates/ignores edges in the network. We explore the proposed method with illustrative examples ranging from static topologies to dynamic ones, considering directed and undirected graphs, presenting several attacking scenarios that are successfully mitigated with our method. Finally, we compare the best undetectable attacking strategy and the commonly used approach named MSR, highlighting the advantages of our method.
Geometric Semantic Genetic Programming with Normalized and Standardized Random Programs
Publication . Bakurov, Illya; Muñoz Contreras, José Manuel; Castelli, Mauro; Rodrigues, Nuno Miguel Duarte; Silva, Sara; Trujillo, Leonardo; Vanneschi, Leonardo; Information Management Research Center (MagIC) - NOVA Information Management School; NOVA Information Management School (NOVA IMS); Springer Science Business Media
Geometric semantic genetic programming (GSGP) represents one of the most promising developments in the area of evolutionary computation (EC) in the last decade. The results achieved by incorporating semantic awareness in the evolutionary process demonstrate the impact that geometric semantic operators have brought to the field of EC. An improvement to the geometric semantic mutation (GSM) operator is proposed, inspired by the results achieved by batch normalization in deep learning. While, in one of its most used versions, GSM relies on the use of the sigmoid function to constrain the semantics of two random programs responsible for perturbing the parent’s semantics, here a different approach is followed, which allows reducing the size of the resulting programs and overcoming the issues associated with the use of the sigmoid function, as commonly done in deep learning. The idea is to consider a single random program and use it to perturb the parent’s semantics only after standardization or normalization. The experimental results demonstrate the suitability of the proposed approach: despite its simplicity, the presented GSM variants outperform standard GSGP on the studied benchmarks, with a difference in terms of performance that is statistically significant. Furthermore, the individuals generated by the new GSM variants are easier to simplify, allowing us to create accurate but significantly smaller solutions.
Exploring SLUG
Publication . Rodrigues, Nuno M.; Batista, João E.; La Cava, William; Vanneschi, Leonardo; Silva, Sara; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School; Springer
We present SLUG, a recent method that uses genetic algorithms as a wrapper for genetic programming and performs feature selection while inducing models. SLUG was shown to be successful on different types of classification tasks, achieving state-of-the-art results on the synthetic datasets produced by GAMETES, a tool for embedding epistatic gene–gene interactions into noisy datasets. SLUG has also been studied and modified to demonstrate that its two elements, wrapper and learner, are the right combination that grants it success. We report these results and test SLUG on an additional six GAMETES datasets of increased difficulty, for a total of four regular and 16 epistatic datasets. Despite its slowness, SLUG achieves the best results and solves all but the most difficult classification tasks. We perform further explorations of its inner dynamics and discover how to improve the feature selection by enriching the communication between wrapper and learner, thus taking the first step toward a new and more powerful SLUG.
An empirical study on the application of KANs for classification
Publication . Costa, Samuel Sampaio; Pato, Matilde; Datia, Nuno; NOVALincs
Kolmogorov-Arnold Networks (KANs) represent a breakthrough in deep learning, diverging from Multi-Layer Perceptrons (MLPs) by generalizing the Kolmogorov-Arnold representation theorem (KAT) to networks of arbitrary depth and width. This theorem facilitates the decomposition of multivariate functions into constituent one-dimensional elements, with learnable activation functions on weights and the sum operator on nodes. KANs have been shown to exhibit robust performance in function approximation, validated across mathematical, physical, and practical domains such as traffic prediction and medical diagnostics. Our study investigates KANs’ efficacy through comprehensive evaluations on OpenML, Kaggle and UCI datasets, with a focus on enhancing Human Activity Recognition systems. They demonstrate high classification performance compared to conventional machine learning approaches and MLPs. These findings underscore KANs’ potential as scalable, interpretable tools in modern machine learning applications given their favorable neural scaling laws.
Slim
Publication . Rosenfeld, Liah; Farinati, Davide; Rasteiro, Diogo; Pietropolli, Gloria; Rebuli, Karina Brotto; Silva, Sara; Vanneschi, Leonardo; NOVA Information Management School (NOVA IMS); Information Management Research Center (MagIC) - NOVA Information Management School
This poster presents Slim: an open-source Python library that provides the first ever framework for the Semantic Learning algorithm based on Inflate and deflate Mutation (SLIM-GSGP). Proposed by Vanneschi in 2024, SLIM-GSGP is a promising non-bloating variant of Geometric Semantic Genetic Programming (GSGP). The Slim library includes all existing SLIM-GSGP variants, as well as traditional GSGP and standard Genetic Programming (GP), facilitating comparative analysis and benchmarking. Additionally, Slim’s semi-modular architecture and parallel computation renders it not only fast but also user-friendly and easily extensible, thereby fostering progress in this emerging area of research.
Unidades organizacionais
Descrição
Palavras-chave
Contribuidores
Financiadores
Entidade financiadora
Fundação para a Ciência e a Tecnologia
Programa de financiamento
6817 - DCRRNI ID
Número da atribuição
UIDP/00408/2020