GA将?開発日記~王理のその先へ~

ネタ勢最強を目指して絶賛開発中。

オーバーフィッティングしてるよ (´・ω・`)

 えー、うちのCritic曰く「三目並べの初期局面の評価値は0.6点」だそうです。勝率80%に相当する数字。

 うん、バグってますね。

 どこが悪いのかな〜。ネットワークがデカすぎる*1? それとも学習率*2? RMSPropが悪さしてるって事は無いよね〜、多分。

15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:先手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:対局中(-1)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==-0, score==0.608975
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |○|  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:後手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:対局中(-1)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==-0, score==0.636282
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |○|×|
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:先手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:対局中(-1)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==1, score==0.712861
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |○|  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |○|×|
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:後手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:対局中(-1)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==1, score==0.691237
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |○|  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |○|×|
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |×|  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:先手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:対局中(-1)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==1, score==0.886164
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() >
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |○|  |  |
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |  |○|×|
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > |×|  |○|
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > +--+--+--+
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > 手番:後手
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > GameState:先手の勝ち(0)
15:14:02 @ core::rl::AcPgleafAgent3::setupGradient() > trueScore==1, score==1

*1:全結合層128ユニット*6層

*2:1E-5