Created by Sashank Pisupati on 9/21/20 for PSY/NEU338: From animal behavior to changing people's minds, taught by Yael Niv
This repository contains a 3-part tutorial on basic reinforcement learning algorithms for across-trial and within-trial associative learning, both pavlovian and instrumental.
- The first part covers Rescorla-Wagner learning, a classic error-based algorithm for associating stimuli with rewards across trials.
- The second part covers Temporal-Difference learning, which allows for learning temporal predictions of value within trials and second-order associations.
- The third part covers three different model-free RL algorithms for learning valuable actions: Actor-critic learning, Q-learning and SARSA.
Each part of the tutorial is a jupyter notebook (can be opened on Colab) containing a brief theory overview and questions with starter code, followed by solutions.