While searching
online I cam across an old e-mail on the computer go mailing list (below). Was
there every any follow up to it? Has anyone made progress in that
regard?
Mark Boon wrote: > I'd rather call it the distinction between global search versus > goal-oriented search. > Goal directed searching. Specifically Hierarchical goal directed search is one thing I'm interested in and have been thinking about posting a question to this list to help design a good hierarchical model of goals for playing go. So this seems like good timing... See "Applying adversarial planning techniques to Go" (http://www.aiai.ed.ac.uk/~johnl/papers/willmott-tcs01.ps) Example start to modeling the goals from higher to lower level objectives: Win (territory + prisoners > opp) Surround Territory Claim territory Corners Sides Center Connect stones Invade territory with hopes to kill or live there Save my groups Limit opponent territory Push into opponent territory to expand my territory Invade territory with hopes to kill or live there Kill opp groups Save my groups Make eyes Play to increase potential eyes Make good eye shape Play on vital eye points Make 2 eyes Connect to living group Connect two weak groups to live Increase liberties Extend/escape Connect to friendly Kill opp groups Squeeze group to limit its space to live Reduce string liberties Put string in atari Double atari Prevent making eyes Play to decrease potential eyes Prevent good eye shape Play on vital eye points Cut kill sub group Please post suggestions about how to best logically model these goals. As you can see I need help improving the logical organization and completeness of the model. It would be great to have the full model that was used in the program Gobi described in the paper above. Or something similar if its available. Matt
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
