Sunday, September 26, 2010

Sea of Yellow

Barke and I had a wonderful time at Jean's, was sure nice of her to let me bring him over and nice of Laverne to drive us. I felt bad that he's home alone all day during the week and didn't want to leave him again on Sat. Jean used to have a German Shepherd and Barke kind of reminds her of him. I ended up winning the day, including playing RA(TI)ONING, a nice 9. In my last round, I bingoed with WANTONER, got a challenge, bingoed again with PRIVATE and then picked up 26 points from the letters on my opponent's rack. Nice way to pull ahead by an extra 200 points.

Made some huge inroads into EII, I've always wanted to do a beam stack search with 100,000 nodes. In english, it means having the computer look at the most promising 100,000 positions before going forward to look at the next ones. In heuristic search, it's common to just look at one idea at a time. With this approach in EII, if you make one bad move early on, you'd have no chance to find the solution. This puts the eggs into 100,000 baskets instead of 1. With a search space of 10^45 and my ability to look at 10^5 at a time, I'm still not holding my breath for the $2 million prize. But with a litmus test of a beam size of 20,000 I was able to find better answers than I have to date. It took it 1 day to run. Now that I have a larger width, 200,000 actually, it should about 10 times as long, so in a week or two I'll see how much farther, if any, it was able to get.

I was able to place 211 / 256 pieces before it determined it would be unsolvable. Not great, but that means all but basically three lines were done.

I could score much higher if I was just counting the matching edges out of 480. But I'm trying to get a full solution, not just a high score for a partial solution.

No comments:

Post a Comment