When dealing with large and complex data, explaining classification decisions can be difficult and hard to interpret, especially when it comes to interactions. This project seeks to combine statistical methodology and variable selection with deep learning to disentangle the complex non-linear interactions present in genomic data. Project P5 aims to develop methods for variable selection and Bayesian regularization that can be applied within the framework of semi-structured mixed models, as well as in models that explain classification decisions such as LRP.
This project will be linked to P4, which focuses on deciphering interactions in genomes using different methodological approaches. We will also explore how to complement our LRP approach with P2’s tests, and we will adapt our Bayesian variable selection techniques to disease progression with neuroimaging data in P6. Additionally, we will share semi-structured regression models with P6 and P7.