Hi guys, Here's my phd thesis (not defended yet): http://www.mimuw.edu.pl/~lew/files/phd.pdf
Most interesting stuff: Section 4.7.6: Simple online variant of Minorization-Maximization with small memory requirements. Section 5.0: Why divide-and-conquer (or equivalently: adaptive-playouts) are important. Section 5.2: Why simple pattern-based adaptive playout can't work well. Section 5.3, 5.4: Why linear model adaptive playout (for instance TD(lambda), Dyna-2) can't work well. Section 5.5, 5.6: Hopefully simple explanation of basic Combinatorial Game Theory and Thermographs. Section 5.8 Monte-Carlo-like algorithm applied to CGT (with simplified thermographs) converges to correct solution on a toy problem (contrary to both previous algorithms). Cheers, Łukasz _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
