Argmax for multilabel classification?

[kckishan]

I wonder why you are considering only the top predicted label out of all possible multiple label predictions. You should check if the model is predicting all labels correctly in case of multilabel classification.

DeepMNE-CNN/code/models.py

Line 203 in 40d1896

true_pred_idx = tf.argmax(tf.multiply(self.scores, self.input_y), 1)

[98764321]

麻烦问一下训练过程损失函数为负，还正常吗？多谢，