Updating stages of alpha are not clear in the code

[mhaj98]

Thank you for the implementation.

I was trying to convert your implementation to be compatible with TensorFlow backend. Comparing the paper with your implementation, I had several issues tracking the updates of the adaptation weights "alphas".

• the "adaptive_weight" configuration is not used. I guess it is also related to updating alphas
• I couldn't track where you are doing the smoothing and normalizing stages of alphas as stated in "Algorithm 1" in your paper.

Thanks in advance for your reply and help.