Explore the intricate relationship between working memory and reinforcement learning in this 50-minute lecture by Anne Collins from UC Berkeley. Delve into computational theories of the brain, examining how dopamine neurons encode reward prediction errors and the dual components of learning optimized for different trade-offs. Analyze experimental results from a study with 78 participants, investigating the role of working memory in learning processes. Discover how genes dissociate prefrontal and N-RL contributions to learning, and why relying solely on RL models can lead to ambiguous results. Examine an EEG experiment that investigates reward prediction errors and learning curves, providing deeper insights into the neural mechanisms underlying these cognitive processes. Gain a comprehensive understanding of how working memory influences reinforcement learning computations in both brain and behavior.
Working Memory Influences Reinforcement Learning Computations in Brain and Behavior
Overview
Syllabus
Intro
Dopamine neurons encode reward prediction error
(at least) Two components for learning: optimized for different trade-offs
Experimental results: n=78
Working memory (WM)
Genes dissociate prefrontal and N-RL contributions to learning.
RL-only model leads to ambiguous results
Are deficits due to N-RL impairment?
EEG experiment
FB-locked: Reward prediction errors?
EEG learning curves
Acknowledgements
Taught by
Simons Institute
