Alumni
Bio
Ian Fan is a final-year student at the University of Melbourne, Australia. He is pursuing a bachelor’s in computer science. He is interested in finding elegant and efficient algorithms to solve problems. The Wolfram Summer School is able to help him to gain experience and lead him to a deeper understanding of computer science.
Computational Essay
Project: Reinforcement Q-Learning for Atari Games
Goal
This project aims to create a neural network agent that plays Atari games. This agent is trained using Q-learning. The agent will not have any a priori knowledge of the game and will be able to learn by playing the game and only being told when it loses.
Main Results in Detail
Successfully implemented a classical Q-learning agent on the CartPole environment and was able to achieve an average performance of 195 episodes in 300 games. Tried to add time information in the input, and the agent achieved 8,000 episodes within 300 games.
Future Work
Make the Q-learning agent with multiple observations as input, make the performance more stable and implement different techniques that increase the network’s performance, like DDQN, NoisyNet, etc.
