Welcome

This repo documents my work to train an agent to solve a Rubik's Cube using a variant of deep reinforcement learning inspired by Playing Atari with Deep Reinforcement Learning (https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf).

Quick Start

To train a neural network to approximate the Q-function run:

python train.py --config_path configs/<config_file>.yaml 

Blog Post

For a more in-depth write-up of the methods used, check out my blog posts on Medium:

Part 1 - https://medium.com/analytics-vidhya/solving-a-rubiks-cube-with-reinforcement-learning-part-1-4f0405dd07f2

Part 2 - https://medium.com/analytics-vidhya/solving-a-rubiks-cube-with-reinforcement-learning-part-2-b4ff0f3522b3