[英]Keras sequential model not training (Stuck on the same Accuracy and Loss)
我正在嘗試為 UCI Tic-Tac-Toe 數據集 ( https://archive.ics.uci.edu/ml/datasets/Tic-Tac-Toe+Endgame ) 構建分類算法,但我遇到了一些問題
Model = Sequential()
Model.add(Dense(9))
Model.add(Dense(64))
Model.add(Dense(64))
Model.add(Dense(1, activation="softmax"))
Model.compile(loss = "binary_crossentropy", optimizer = "Adam", metrics = ["accuracy"])
Model.fit(X_Train, Y_Train, batch_size = BATCH_SIZE, epochs = EPOCHS, validation_data = (X_Val, Y_Val))
我在所有 Epochs 中都收到了這條消息
Epoch 100/100
861/861 [==============================] - 0s 40us/step - loss: 5.3782 - accuracy: 0.6492 -
val_loss: 4.7916 - val_accuracy: 0.6875
有誰知道解決這個問題的方法
你不能對一個神經元使用 softmax,如果它是二元分類,你應該在輸出層使用sigmoid
激活:
Model.add(Dense(1, activation="sigmoid"))
聲明:本站的技術帖子網頁,遵循CC BY-SA 4.0協議,如果您需要轉載,請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.