Bayesian Pattern Ranking for Move Prediction in the Game of Go by David Stern, Ralf Herbrich and Thore Graepel at Cambridge / Microsoft uses nested patterns with learnt priority (rank) to "obtain a probability distribution for professional play over legal moves in a given position." Trained using a corpus of 20000 professional games they also discuss building a million game corpus of variable-level games (presumably games played on internet go servers?).
It seems to me that the problem they're setting out to solve isn't the one they need to solve―at the end of the day generating reasonable moves is the goal, and pro moves are only an approximation. There is also the problem that the system assumes that the board positions and the opponents' moves are reasonable. There were some clear attacks against early computer chess players which worked by playing unreasonable moves to get out of the opening book.
Another issue is the shape of he pattern. The patterns are (effectively) circular, whereas there are many, many, go proverbs suggesting that they should be biased towards the board edge (where territory is to be made) over the center.
3 comments:
The patterns used are actually very sensitive to the edge despite being circular in shape. The reason is that the edge, or rather the off-board region, is explicitly represented in the patterns, which match values black, white, empty or off-board for every vertex in the template. A typical pattern thus includes part of the edge or even of two edges (in the corner) and hence the system can take the particular value of the edges into account. For smaller patterns, we are now working on including a feature "distance to edge" as used, e.g., in van der Werf et al's "Local move prediction in Go" paper
Hello, I do not agree with the previous commentator - not so simple
Thank you very much on your pleasant pointers you’ve shared on this page.
Need help on billing? Please visit:
medical billing services in Atlanta
The best billing software for medical professionals
Post a Comment