Visual Analysis of Feature Spaces for Explainability and Enhancement of Distributed Machine Learning

• Grant number: 25/23139-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: January 01, 2026
• End date: December 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Danilo Medeiros Eler
• Grantee: Enzo Nicolás Stromberg Racciatti
• Host Institution: Faculdade de Ciências e Tecnologia (FCT). Universidade Estadual Paulista (UNESP). Campus de Presidente Prudente. Presidente Prudente , SP, Brazil

Abstract

Advances in Deep Learning have enabled applications in critical domains, yet challenges remain regarding model explainability and data quality, particularly in distributed learning settings such as the travelling model paradigm. In these scenarios, decentralization and data confidentiality hinder direct control over training data, limiting process interpretability. This project proposes a methodology for the visual analysis of feature spaces from Convolutional Neural Networks to provide explainable support for data quality management in distributed learning. We hypothesize that visualizing the model's internal representations can reveal patterns indicating the positive or negative influence of local data on overall performance. The expected outcome is an interactive analysis framework that complements traditional metrics with visual insights, enhancing transparency, reliability, and technical rigor in the application of Deep Learning to critical, particularly medical, domains. (AU)