CAIMS Code of Conduct
“All participants are expected to comply with the CAIMS-SCMAI Code of Conduct”
Plenary Lectures
Learning dynamics and generalization in artificial and biological neural networks: a common mathematical perspective
In this talk, I will present an overview of recent advancements in deep learning theory that characterize learning dynamics in large network parameter spaces. First, I will discuss how generalization —the capacity of neural networks to solve tasks on examples that were not seen during training— is influenced by distinct informative features in training data that are learned at different rates. Second, I will present recent results leveraging deep learning theory to understand learning in biological neural networks, where loss landscape curvature and biologically plausible optimization combine to influence generalization properties.