[Computer-go] A Linear Classifier Outperforms UCT on 9x9 Go

Lucas, Simon M sml at essex.ac.uk
Wed Jun 29 15:42:23 PDT 2011


>Such a table has 47 million entries in 19x19 go and over half a million in 9x9 go.   It require an enormous amount of data to fill such a table with enough samples to not be almost meaningless statistically.


 Maybe I'm missing something in the context of this statement, but I don't think the size
 of the table is necessarily a problem when considering the amount of data
 required to ensure the entries are statistically meaningful: it depends so much
 on the distribution of the data.   For example, it sounds like those sizes are based
 on all the entries that could possibly occur: perhaps in practice only a relatively small
 number of entries actually occur, and the statistics of those occurrences can be
 estimated reliably (reliably enough to add value) from a relatively small number of games (similar things happen
 when using n-grams for statistical language modelling).

 Simon Lucas

http://csee.essex.ac.uk/staff/lucas/






More information about the Computer-go mailing list