Creating A More Explainable and Efficient Game-Playing Agent
February 04, 2022
news-image

Wednesday, February 9th at 1pm, Summit Room 3346 TMCB

Advisor: Chris Archibald

MS Thesis Proposal for Jamison Moody

Abstract:

Why does it take us humans only a few minutes to learn how to play a game at a basic level but it takes machines much longer? The answer is that we have prior knowledge of objects, and we can abstract the world in a way that allows us to react to a situation we have never seen before. With written instructions, we are able to perform tasks we have never seen another human do. For my thesis, I plan to explore how to incorporate prior knowledge (knowledge of objects and natural language understanding) with Reinforcement Learning. The eventual goal is to create a more efficient and explainable learning agent.