DQN Cartppole-V1 にてTensorboardを使って学習をグラフ化しました

上記のグラフのLossの下りと同時にスコアは改善されてます：

episode: 0   score: 10.0   memory length: 11   epsilon: 1
episode: 5   score: 76.0   memory length: 154   epsilon: 1
episode: 10   score: 16.0   memory length: 278   epsilon: 1
episode: 15   score: 17.0   memory length: 364   epsilon: 1
episode: 20   score: 17.0   memory length: 491   epsilon: 1
episode: 25   score: 32.0   memory length: 596   epsilon: 1
episode: 30   score: 12.0   memory length: 657   epsilon: 1
episode: 35   score: 20.0   memory length: 742   epsilon: 1
episode: 40   score: 16.0   memory length: 846   epsilon: 1
episode: 45   score: 19.0   memory length: 954   epsilon: 1


self.epsilon 0.999
episode: 50   score: 19.0   memory length: 1126   epsilon: 0.8806777104745716
episode: 55   score: 9.0   memory length: 1238   epsilon: 0.7873207291459607
episode: 60   score: 17.0   memory length: 1389   epsilon: 0.6769247732130653
episode: 65   score: 14.0   memory length: 1545   epsilon: 0.5791040088995179
episode: 70   score: 65.0   memory length: 1821   epsilon: 0.4393709323780249
episode: 75   score: 60.0   memory length: 2000   epsilon: 0.3104958044435009
episode: 80   score: 119.0   memory length: 2000   epsilon: 0.20622457658762192
episode: 85   score: 201.0   memory length: 2000   epsilon: 0.07402874109670564
episode: 90   score: 239.0   memory length: 2000   epsilon: 0.02485123742451863
episode: 95   score: 284.0   memory length: 2000   epsilon: 0.009998671593271896
episode: 100   score: 202.0   memory length: 2000   epsilon: 0.009998671593271896
episode: 105   score: 292.0   memory length: 2000   epsilon: 0.009998671593271896

Tensorboardの使い方はこちらが分かりやすいです：https://stackoverflow.com/questions/42112260/how-do-i-use-the-tensorboard-callback-of-keras

以上でした。