Study and Implementation of Techniques for Explaining Predictive Models

• Grant number: 25/01862-3
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: May 01, 2025
• End date: April 30, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Marcos Gonçalves Quiles
• Grantee: Jonas Lucas Durão
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil
• Company: Universidade de São Paulo (USP). Instituto de Química de São Carlos (IQSC)
• Associated research grant: 17/11631-2 - CINE: computational materials design based on atomistic simulations, meso-scale, multi-physics, and artificial intelligence for energy applications, AP.PCPE

Abstract

Explainable Artificial Intelligence (XAI) has gained prominence for providing transparency and understanding in the decision-making processes of machine learning models. This explainability is crucial in scientific areas, where trust in predictions must be high and decisions based on these models can have significant implications. In the field of chemistry, the use of XAI is particularly relevant. Machine learning models have shown promise in performing various tasks, as opposed to traditional approaches that demand high computational costs, such as Density Functional Theory (DFT) calculations. The ability to interpret and understand the predictions of these models is vital for domain experts, allowing them greater confidence in the recommendations and more effective validation of the obtained results. Despite its importance, there is a significant lack of studies integrating XAI with the prediction of material properties, highlighting a critical gap in this research field. This research project aims to fill that gap by focusing on the interpretation of molecular property predictors based on neural networks, which are considered black-box models. The central objective is to build predictive models that not only achieve high accuracy in the prediction of material properties but also provide clear and understandable insights into how these predictions are made, strengthening the trust and applicability of machine learning models in materials chemistry. In particular, the student will study and implement various XAI methods and make them available in a Python-based toolbox.