AI (Reinforcement Learning)

A summer research program, devised an RL model based on the A3C algorithm and other unique techniques, defeated low-difficulty robots in SC2.

Role: Lead Programmer
Duration: Summer 2019
Tool: Python, TensorFlow, pysc2

This was a summer research program, a team of 4. We devised a model able to defeat low-difficulty robot agents based on the A3C algorithm and unique methods, such as combining multiple action functions into a single task, filtering out unnecessary inputs, and setting the focus in battle. I took charge of the algorithm design and contributed most of the code to the project as a lead programmer.

↑Back to Top