Task-Induced Representational Invariances Depend on Learning Objective in Deep RL
Reinforcement Learning (RL) has long served as a model for goal-directed animal behavior in neuroscience. Modern deep RL has shown…
arXiv:2606.01868NeuroAI and Geometric Data Analysis
Our group develops mathematical theories for understanding how neurons collectively give rise to behavior in biological and artificial neural networks. Our current focus is on addressing this question through two broad approaches at the intersection of computational neuroscience and deep learning:
(1) analyzing geometries underlying neural or feature representations, embedding and transferring information, and (2) building neural network models and learning rules guided by neuroscience. To do this, we combine computational tools from theoretical physics, applied math, and machine learning. Alongside this theoretical work, we develop close collaborations with experimentalists to be inspired by and to test ideas on neural data.
NeuroAI & Geometric Data Analysis Lab
Current Papers
Two Speeds of Learning: A Representation-Readout Decomposition of Grokking and Double Descent
Training loss and accuracy are the standard signals used to monitor generalization during deep neural network training. Two well-documented phenomena…
arXiv:2605.27078v2Estimating Dimensionality of Neural Representations from Finite Samples
C. Chun, A. Canatar, S. Chung , Daniel Lee
The global dimensionality of a neural representation manifold provides rich insight into the computational process underlying both artificial and biological…
International Conference on Learning Representations (ICLR)
