Today was one busy day. I learned a lot of things.
- I learned about Propositional net (PropNet) and attempts to build a general game player using PropNet. Wow this is hard. We (my teammates and I) spent roughly 5 hours on it and didn’t get that far. We followed the pseudocode in the lecture notes for the GGP class at Stanford but the pseudocode isn’t that clear. We will have to spend more time on it tomorrow. When we finish, I hope to write a detailed account of our experiences with General Game Playing and PropNet. On a slightly related note, our Monte Carlo Tree Search player is a beast. We have been consistently winning at the game Connect Four.
- I built my first recurrent neural network last weekend. A month ago I didn’t even know what a neural network was and now I have built both a deep neural network (more like a shallow one with only 1 hidden layer) and a recurrent neural network. It’s incredible how far a class can take you. I can write about my experiences with neural network over the summer too.
- I made a list of things I want to achieve within the next year, and I realized that I’d better work twice as hard if I want to achieve any of them. Exciting time ahead!