About

About Website

This website is a devlog for my personal project with the goal of letting Yonaka play games.

I will be using the posts as a way to log my experiments and experience, and methods I develop while working on the project.

The project will mostly consist of working with Reinforcement Learning, specifically Q learning instead of any of the “advanced” policy gradient methods.

This is because I believe value based methods are more extendable and higher potential, once you properly understand them. That is why I’m starting by extending Q learning to continuous actions and multiple actions first, as those is seen as the major limiting factor of Q learning.