Un (petit) pas vers l’explicabilité de l’intelligence artificielle
Romain Giot · HAL (Le Centre pour la Communication Scientifique Directe) · 2024
This document outlines my contributions since joining the LaBRI in September 2013, and my research plans for the coming years. The primary areas of research I discuss are:- Data analysis and machine learning. Here, I present work on biometric authentication and how features are attributed within deep neural networks.- Information visualization. This involves my research on techniques for drawing large graphs and grid-based visualization methods.- The intersection of information visualization and machine learning. I describe research on using visualization to aid machine learning, including biometric authentication, analysis and visualization of labeled datasets, and enhancing the understanding of deep neural networks. Additionally, I discuss using machine learning to improve visualization, such as automatically comparing visualization techniques, improving graph drawing, and removing overlaps in graph visualization. Research perspectives focus on this aspect.