I figure it’s been a while since I’ve blogged here, and I’ve been feeling like I needed to exercise my mind some. Here’s a little something I’ve been thinking about:
A man was inside my house today. Not that he was let in, he let himself in. And as I stood in the doorway [...]

Ran a modified consoleTrainer yesterday and found some results for all domains of Tetris.
More to come: I have class. Note that parameter 6 is particularly nasty.
Back. Here are the results for the various parameters (Note that the / is used to separate the runs):
Initial run / eTemp’d run / BETA = 10 / BETA = [...]

Ergh. Both of the tests running from Friday were writing to the same file so I don’t know which one is the end file. Have to run it again…
Presentation complete. Just need to present it now.

After running a ‘learn the best’ run (set eTemp to 0.5 with no cooling), the results may be a bit more realistic, even if they look disappointing. The problem lies in the fact that the other fields are nearly all smaller or have a less balanced proportion of pieces than the ‘fair’ first MDP which [...]

Although the eTemp wasn’t being reset in the agent_init method, it wouldn’t have mattered in this case. In the proving run though, it would have been quite bad.
Anyway, the results obtained appear much the same, so clearly the rate of learning is the problem.
There are three different areas to change: The Beta value (multiplier of [...]