Workshop on Reinforcement Learning

Co-organized and delivered short lecture on reinforcement learning.

Description:

A group of 16 participants, which included students and researchers from diverse fields, such as 9 from CASCB, 3 from Biology, 3 from Physics, and 1 from Psychology, attended a two-day deep reinforcement learning workshop. The first day was spent discussing key concepts and algorithms in reinforcement learning. Participants learned the differences between states, actions, policies, value functions, and several value estimation methods.

Participants completed coding tasks using Google Collab, which taught them about the algorithms for value estimation. On the second day, we covered more advanced topics like Q-learning and Deep Q-learning. The session involved experimenting with Chase and Escape agents, which is a fundamental concept of multi-agent predator and prey. At the end of the workshop, the participants formed three groups to brainstorm ideas for using reinforcement learning in their own research. We got three main topics: investigating the emergence of collective cell movement, simulating the movement of bacteria based on light gradients, and exploring whether theory of mind might emerge from interacting agents within a dominance hierarchy. Participants left the workshop with a general understanding of reinforcement learning methods, and how to approach applying them to their own questions.