Study of Interpretable Deep Neural Networks in Brain-Computer Interfaces

• Grant number: 24/07783-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: August 01, 2024
• End date: July 31, 2025
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Levy Boccato
• Grantee: Helena Guachalla de Andrade
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação (FEEC)
• Associated research grant: 20/09838-0 - BI0S - Brazilian Institute of Data Science, AP.PCPE

Abstract

The application of deep neural networks in brain-computer interfaces (BCIs) has intensified, mainly due to these models' potential to analyze raw electroencephalogram (EEG) signals. In particular, convolutional neural networks (CNNs), which have excelled in image classification and speech recognition, have also been explored.In this context, using interpretable architectures that perform well in the direct processing of EEGs becomes desirable. By understanding how the model makes decisions, as well as which data attributes were decisive for the obtained result, it is possible to increase reliability and robustness. Specifically, two well-established proposals that explore this characteristic are the SincNet and EEGNet networks. This project aims to study different interpretable neural network architectures developed for the treatment of temporal signals and to systematically apply these approaches in the context of EEG-based BCI systems that explore different paradigms. Hence, we hope to provide a comprehensive overview of the potentialities and limitations of each technique and discuss the relevance of interpretable elements in designing more robust BCIs.