# 覺得哪裡怪怪的請趕快跟助教說
for 二分類：
1. "output" layer: units=1, activation='linear'
2. loss = tf.keras.losses.BinaryCrossentropy(from_logits=True))
3. prediction = (sigmoid(model.predict(X_te)) > 0.5).astype( "int32" )
4. validation: acc, recall, precision, ...
附件列表