Myersna-log

These are my (semi) daily logs for my project. You can find my project summary page here

November 29

I've spent most of today thinking about the actual implementation of the genetic algorithm, and laid out the stubs and extensions that are needed to retrofit my minmax to become genetic. Hopefully I got all the math right.

November 26

Met with Jim, laid out a fairly exhaustive species representation for the game. Still need to get the dumps from minmax, should be easy but didn't get around to it today. My goal this week is to clear out all of my other work by Friday so that I can spend the remainder of the semester more or less focusing solely on my project.

November 25

Back from break. Over break I thought a lot about ways to express policies for the game genetically, I'm still working on a way to condense the information down so that meaningful progress is more likely to happen during cross-over and mutation. For the fitness function I decided it would be best to use a static policy that plays within random parameters. For example a policy that always plays a card within 10% of the value of the flop (if possible). Still, the biggest hurdle is creating the blueprints for the species.

November 13

I've written up a crude Gant(?) chart stating what I need to do, first I need to get this minmax working correctly with output on the double. I've been spending probably too much time on other projects. Also at the same time I need to look more into data formats that are compatible with genetic algorithms.

November 7

• Got an implementation of minmax working, except that it's not minmax. Going to change the structure to relfect minmax.

November 3

• Continued work on Minmax
• Created my project page

October 31

Switched to a Wiki format from text file

October 30

Continued work on Minmax

October 29

Began work coding Minmax, basic framework laid out Began outline of Paper.

October 28

Looked a bit at the other students logs, mine looks really bad in comparison...I need to get this to wrap properly.

October 27

Continued reading new research, comparitivly I'm finding the chess based readings a little dry.

October 26

Read up on some research conducted yesterday, An interesting article regarding believable AI in games found (VS actually intelligent AI). Talked about giving AI human-like limitation. Got me thinking about whether it would be possible to train a learning AI about a game merely by giving it large amounts of replay data from games and letting it reason strategies and tactics by matching already seen environments. This is counter to letting the AI actually play out the game. Perhaps I'm just talking about training data and not relizing it.

October 25

Did some additional research that is more focused on what I want to do, finding research on AI and machine learning as applied to strategy games. Most of them pertain to chess, in a non-brute force method.