ニューラルネットワーク ReLu関数でエポック毎の結果がおかしくなってしまう
解決したいこと
ReLu関数でエポック毎の結果がおかしくなってしまう
こちらのサイトの以下の記事でニューラルネットワークのタイタニックデータを使ったフルスクラッチを学習させて頂いてます。
以下の記事では隠れ層の活性化関数にシグモイド関数を使用してますが、ReLu関数に変更して実装したいのですが、エポック毎の精度が全て同じになってしまいます。
初心者なので何をどうすれば解決するかが分からない状態です。
手作り3層ニューラルネットワークをフルスクラッチで実装して、 Kaggle Titanic コンペに Submit してみた
https://qiita.com/nozomale/items/197243963e1e96d42680
発生している問題・エラー
シグモイド関数の代わりに以下のReLu関数を入れ替えて実装したのですが、
def relu(x):
return np.maximum(0, x)
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264
train acc : 0.615 || test acc : 0.6263736263736264