description | determine playing style in the game of go |
owner | pasky@ucw.cz |
last change | Tue, 23 Jul 2013 18:39:12 +0000 (23 20:39 +0200) |
URL | git://repo.or.cz/gostyle.git |
| https://repo.or.cz/gostyle.git |
push URL | ssh://repo.or.cz/gostyle.git |
| https://repo.or.cz/gostyle.git (learn more) |
bundle info | gostyle.git downloadable bundles |
content tags
|
|
README
Hello!!
This repository includes various tools used to analyze Go-players style. Whats here:
- a tool for database of Go games, whose data structures (like a list of games)
we use all over the place, read utils/godb_models.py and utils/godb_session.py
- a simple wrapper for pachi to work with the games and scan the raw patterns comfortably
(see pachi.py)
- stuff to work and form pattern vectors from raw patterns (see game_to_vec.py )
and to export the vectors as .tab files which can be then used by the Orange Datamining
toolkit to fiddle around with some machine learning stuff, http://orange.biolab.si/
All the major files have examples at the bottom and are pretty well commented.
REQUIREMENTS:
- if you want to scan the games, you need pachi go engine, instructions below
- if you want to toy with machine learning stuff, you need the Orange datamining toolkit
get it from http://orange.biolab.si/
------
How to prepare pachi:
run this in some dir (assuming in a parent dir of gostyle - this dir)
$ git clone git://repo.or.cz/pachi.git
$ cd pachi
$ git checkout jmhack
# this is a ugly way to make the contiguity feature say the real distance
# instead of a proximity flag it does by def.
$ patch pattern.c ../gostyle/PATCH_pachi.diff
$ make
# this is necessary, otw pachi segfaults
$ touch patterns.spat