What is the difference between binary cross-entropy and binary cross-entropy with logits in keras?

Question

In keras backend we have a flag with_logitsin K.binary_crossentropy. What is the difference between normal binary crossentropy and binary crossentropy with logits? Suppose I use the seq2seq model and my output sequence is of type 100111100011101.

What should I use for recursive LSTM or RNN to learn from this data? Do I provide a similar input sequence along with timesteps?

Answer 1

It depends on whether you have a sigmoid layer immediately before the loss function.

, , from_logits False. , tf.nn.sigmoid_cross_entropy_with_logits.

(.. ), from_logits=True, .

, . , logit , , .