vanhavel/bp-mll-tensorflow

Implementation of the BP-MLL loss function in Tensorflow

bp-mll-tensorflow's Issues

Does this work with CNN?

I am running the code here: https://colab.research.google.com/drive/1Da16UW8PudKFfGs756vClp0G_IpiJsGA

But after some steps the loss is NaN.
Is there something I am doing wrong?

Full example in keras

Do you have a model definition that works in keras?

I'm having problemas because loss is becoming NaN , after 10 steps. I don't know if it's the gradients or anything like this.

Activation of output layer

Hi,

Just noticed in the paper activation of the output layer is tanh while in your example you used sigmoid.

Is it possible to use both?