Equivariant DQN in CartPole

Project Overview

This project focused on applying Deep Q-Networks (DQN) to solve the CartPole problem in the OpenAI Gym, a benchmark task in reinforcement learning. By leveraging equivariant model techniques, the project aimed to enhance the efficiency of training DQNs using both state and image inputs.

Achievements

• Equivariant Model Techniques: Applied advanced equivariant model techniques to improve the generalization of the DQN, enabling it to learn more efficiently from the environment.
• 2x Training Speed: Successfully achieved a twofold increase in training speed compared to conventional DQN approaches, significantly reducing the time required to reach optimal performance.