Fuzzy and Probabilistic Logics with Applications to Neural Network Interpretability

• Grant number: 24/19144-7
• Support Opportunities: Regular Research Grants
• Start date: February 01, 2025
• End date: January 31, 2027
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Mobility Program: SPRINT - Projetos de pesquisa - Mobilidade
• Principal Investigator: Marcelo Finger
• Grantee: Marcelo Finger
• Principal researcher abroad: Felip Manyà
• Institution abroad: Consejo Superior de Investigaciones Científicas (CSIC), Spain
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated researchers: Sandro Márcio da Silva Preto
• Associated research grant: 19/07665-4 - Center for Artificial Intelligence, AP.eScience.CPE

Abstract

Neural networks are behind many of the most advanced intelligent systems, yet theirlack of interpretability and explainability remains a significant challenge. In a neurosymbolic approach, the infinite-valued Lukasiewicz logic ( L) may be used to representneural networks in a formal language, enabling formal verification of their propertiesthrough logical formula evaluation. While previous efforts have focused on plain L,another possibility is to leverage FP( L), a probabilistic extension of L, which is bettersuited for handling the probabilistic nature of neural networks. The main objectivesof this project are to advance the computational and algorithmic understanding ofFP( L). Furthermore, this project seeks to explore future collaborative opportunitiesand propose new joint projects between researchers from São Paulo and Barcelonawith an emphasis on improving neural network interpretability through logical systems,along with other topics in artificial intelligence. (AU)