HCha/Stockfish
1
0
Fork 0
mirror of https://github.com/HChaZZY/Stockfish.git synced 2025-12-25 11:36:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into master

Browse Source
This commit is contained in:
nodchip
2020-06-28 11:42:23 +09:00
committed by GitHub
parent aea08de018 f1a8580118
commit 123dd68452
4 changed files with 22 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
src/learn/learner.cpp
View File

@@ -1028,24 +1028,10 @@ double sigmoid(double x)
// A function that converts the evaluation value to the winning rate [0,1]
double winning_percentage(double value)
{
// In Maxima,
// load("C:/maxima-5.44.0/cform.lisp");
// PawnValueEg = 206;
// cform(1.0 / (1.0 + 10.0 ^ (-value / PawnValueEg / 4.0)));
constexpr double PawnValue = PawnValueEg;
return 1.0 * pow(pow(10.0, -0.25 * pow(PawnValue, -1) * value) + 1.0, -1);
}
double delta_winning_percentage(double value)
{
// In Maxima,
// load("C:/maxima-5.44.0/cform.lisp");
// PawnValueEg = 206;
// cform(diff(1.0/(1.0+10.0^(-value/PawnValue/4.0)),value));
constexpr double PawnValue = PawnValueEg;
return
0.5756462732485115 * pow(PawnValue, -1) * pow(10.0, -0.25 * pow(PawnValue, -1) * value) *
pow(pow(10.0, -0.25 * pow(PawnValue, -1) * value) + 1.0, -2);
// 1/(1+10^(-Eval/4))
// = 1/(1+e^(-Eval/4*ln(10))
// = sigmoid(Eval/4*ln(10))
return sigmoid(value / PawnValueEg / 4.0 * log(10.0));
}
double dsigmoid(double x)
{
@@ -1143,7 +1129,6 @@ double calc_grad(Value deep, Value shallow , const PackedSfenValue& psv)
const double q = winning_percentage(shallow);
const double p = winning_percentage(deep);
const double dq = delta_winning_percentage(shallow);
// Use 1 as the correction term if the expected win rate is 1, 0 if you lose, and 0.5 if you draw.
// game_result = 1,0,-1 so add 1 and divide by 2.
@@ -1154,9 +1139,7 @@ double calc_grad(Value deep, Value shallow , const PackedSfenValue& psv)
// Use the actual win rate as a correction term.
// This is the idea of ​​elmo (WCSC27), modern O-parts.
const double pp = (q - p) * dq / q / (1.0 - q);
const double tt = (q - t) * dq / q / (1.0 - q);
const double grad = lambda * pp + (1.0 - lambda) * tt;
const double grad = lambda * (q - p) + (1.0 - lambda) * (q - t);
return grad;
}