Yoel Sanchez Araujo is a first generation Latino Ph.D. student at the Princeton Neuroscience Institute. His main research interests include: using and developing statistical methods for describing neural data, reinforcement learning and broadly the topic of learning and decision making at the systems level. He’s currently formally advised by Professors Jonathan Pillow and Nathaniel Daw, and collaborates with Professor Ilana Witten. He’s additionally interested in programming languages and broad topics in mathematics.
Principal Investigator: Jonathan Pillow
Fellow: Mehnoor Khan
“Characterizing reinforcement learning algorithms using neural data”
The overall goal is for the fellow to gain experience and knowledge of reinforcement learning, model comparison, and how to generally preprocess and handle neural data. The fellow will explore two different reinforcement learning (RL) algorithms: one that performs policy iteration only and another that performs both policy and value iteration. Intuitively, algorithms that perform both policy and value iteration (e.g., actor-critic) seem better suited to explaining animal learning behavior in laboratory settings. The goal of this project is to assess this idea quantitatively. The fellow will fit RL algorithms to neural data and examine how well they account for neural dynamics. In doing so, they will learn techniques for model comparison and model evaluation, such as cross validation, Bayes factors and information criteria for model selection (e.g., AIC).