Deep Reinforcement Learning Hands-On - Second Edition

24

RL in Discrete Optimization

Next, we will explore the new field in reinforcement learning (RL) application: discrete optimization problems, which will be showcased using the famous Rubik's Cube puzzle.

In this chapter, we will:

• Briefly discuss the basics of discrete optimization
• Cover step by step the paper called Solving the Rubik's Cube Without Human Knowledge, by UCI researchers Stephen McAleer et al., 2018, arxiv: 1805.07470, which applies RL methods to the Rubik's Cube optimization problem
• Explore experiments that I've done in an attempt to reproduce the paper's results and directions for future method improvement

RL's reputation

The perception of deep RL is that it is a tool to be used mostly for game playing. This is not surprising ...