[computer-go] Scalability study suggestion
Mogo_16 is about 80 Elo points stronger than Mogo_15 then we could pretend that Mogo_16 has 2 wins and 1 loss against Mogo_15 before any actual games are played. This would seed Mogo_16 into about the right rank, and it would immediately start playing high information games instead of games against opponents that are too weak to yield much information. Chuck Paulson www.PuffinwareLLC.com http://www.puffinwarellc.com/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Re: euler numbers
Here are some links I found while reading the Euler numbers thread. The first one is mentioned by Don and looks very useful for building an actual algorithm. Its A Novel Morphological Operator To Calculate Euler Number by Zhang and Stoecker. http://scholarsmine.umr.edu/post_prints/pdf/00930291_09007dcc8030c7df.pdf Next is the Gray paper Local Properties of Binary Images in Two Dimensions. http://turing.iimas.unam.mx/~elena/CompVis/Gray71BinaryImages.pdf This paper has a brief description of the Quad algorithm applied to a game called Lines of Action (LOA). http://www.cs.unimaas.nl/m.winands/documents/The_Quad_Heuristic_in_Lines_of_ Action.pdf Finally, there is this brief paper by my old undergraduate adviser at MIT that extends the Euler number calculation to continuous gray scale domains, but also briefly mentions the discrete case. http://people.csail.mit.edu/bkph/articles/APE_continuous.pdf Chuck Paulson www.PuffinwareLLC.com http://www.puffinwarellc.com/ iMetaSearch - indexing and clustering search results with Latent Semantic Analysis ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Language
My 2 cents about languages. C is the universal assembly language. I don't think I've ever used a computer family that didn't have a C compiler on it (after C was invented of course). Often new languages, to get started, will just translate into C code and then compile with the C compiler. I wrote my first Go programs earlier this year. I first used Ruby and it was short and easy to write. The GTP protocol (enough for CGOS) took only about 1 page of code. However in timing tests, it could only do about 30 game simulations per second. This was unacceptable and I abandoned Ruby. Next I translated the ideas into C++. Everything was more work, but I anticipated a 10-20 times speed up so it seemed the tradeoff would be worth it. After finishing, I did the same timing tests as with Ruby and it did 9000 game simulations per second without much optimization. I knew, of course, that Ruby is slower than C++ but a factor of 300 is amazing. It helps to have explicit control of memory and mature C compilers that generate fast code. I am still wondering how Ruby could be so much slower than C++. Perhaps this problem is just not suited for Ruby. Chuck Paulson ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
