From d1aa30b1554f4e62748342423b53d3546fc36367 Mon Sep 17 00:00:00 2001 From: Petr Baudis Date: Sat, 6 Mar 2010 19:03:31 +0100 Subject: [PATCH] tex: Elaborate somewhat on the extracted patterns --- tex/gostyle.tex | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-) diff --git a/tex/gostyle.tex b/tex/gostyle.tex index aba193e..1fe5768 100644 --- a/tex/gostyle.tex +++ b/tex/gostyle.tex @@ -375,14 +375,28 @@ for finding correlations between moves of players of the same strength rank. In order to generate the required compact description of most played moves, for each player, we extract a~generic description from each move -played by the player and assign the player a~vector $\vec p$ describing relative -frequency of moves with the same description. These vectors are called \emph{pattern vectors} -in the text. - -TODO: Elaborate, exact construct. +played by the player, then take the most occuring $n$ patterns across all players% +\footnote{We use $n=500$ in our analysis.} and assign each player a~{\em pattern vector} +$\vec d$ where each dimension corresponds to the number of occurences of +one given pattern normalized to range $[0,1]$. \subsection{Pattern Features} +Of course a big question is how to compose the pattern descriptions. +There are some tradeoffs in play - overly general descriptions carry too few +information to discern various player attributes; too specific descriptions +gather too few specimen over the games and the differences in vectors are +not statistically significant. + +We have chosen an intuitive and simple approach inspired by pattern features +used when computing ELO ratings for candidate patterns in Computer Go play. +\cite{ELO} Each pattern is combination of several {\em pattern features} +matched at the position of the played move. We use these features: + +\begin{itemize} +\item TODO +\end{itemize} + \subsection{Expert-based knowledge} \label{style-vectors} Apart from the pattern vectors extracted from game collections, we have gathered some -- 2.10.5.GIT