Keras.backend necessity


I see some times tf.keras.backend is used in the labs. why do we use it? why don’t we use abs and other functions from tf itself.

my main question is what the keras.backend is basically.

for example when we want to define RMSE, we use this. we rather can use tf.abs and … .


here is a nice article I found about Keras and Tensorflow on google: Keras - TF and this one too is nice one Keras Backend

Thanks a 1M :slightly_smiling_face:. your links are great and too much informative. when I wanted to ask this question I thought it would a too much easy one. but when I now see your answer, I see there is a lot of nice and exciting information behind the “backend”.

Anyway for short, referring to your second link…
"If you want the Keras modules you write to be compatible with both Theano and TensorFlow, you have to write them via the abstract Keras backend API. "

1 Like