Keynote
Dr. Michel Besserve
15:00 27 July 2022
Max Planck Institute for Intelligent Systems, Tübingen | Research Group Leader | Website
Can machine learning close the gap between brain data and neural mechanisms?
The vast range of methodologies offered by machine learning to process highly multivariate data holds promises for experimental sciences. However, finding the right tool to address a specific question if often challenging. This is particularly the case in the context of systems neuroscience, where the mechanisms that support brain functions remain largely elusive. Why do we sleep? How do we remember? What is consciousness? These high-level questions relate in non-trivial ways to (1) the organization of neural networks at multiple spatial and temporal scales, (2) the data recorded by experimentalists. We will go over these topics using a range of multivariate analysis techniques, focusing on unsupervised learning, dimensionality reduction and causal inference. Finally, we will elaborate on the limitations of current approaches for producing scientific explanations, and see how recent progress in causal representation learning may unlock a mechanistic understanding of systems despite their complexity.
