As part of the Canadian Distributed Mentorship Program, here is where I will be posting my journal entries relating to the work done.

This week I started reading Sutton and Barto's Reinforcement Learning book, in particular chapters 1-8. What's news to me is that I am actually enjoying the reading. I haven't had a truly interesting textbook since my first year political science class, and the only reason that one was interesting was because it wasn't linear algebra. Yay! To make sure I was on track, I implemented both a Policy Evaluation, as well as a Sarsa algorithm. The former is rather simple, and involves an agent interacting with a simple gridworld environment, in an attempt to reach a target. I implemented the Sarsa algorithm as part of a maze environment, with the learning agent again trying to optimally reach a target point. Perez's Maze Learning example was very useful.

Doina asked me to port the Cat and Mouse environment, originally developed by Eden, Knittel and van Uffelen to work over a network, as well as be compatible with an environment for reinforcement learning agents developed at McGill. I've started working on it, but I have no interesting results to show. Gone are the days when code was written in a day.

I've also decided it was about time to start working on this webpage. Since I am positively HTML-inept, my old friend from high school Jun Esmores, has agreed to take on the rather annoying task of helping me. If there would be such thing as HTML baby sitting, I'm sure this would be it :)