vshmyhlo / machine-learning-playground Goto Github PK

I can see tf.contrib.seq2seq.hardmax being used in the run notebook

2 questions

1. What is the theoretical underpinning behind using Hardmax operator over other options?
2. Why is it only used in the First LSTM layer (out of 3)
3. Does this notebook implement some paper? If yes, could you point to that? If no, what's the model architecture and reasoning?
4. Could you comment the notebook so that it would be make it easier to understand.

This notebook works and is pretty awesome. But just wanted to comprehend it properly. Good job!
Thanks.

@v-shmyhlo Any help in this regard would be appreciated

Could you point to what Datasets folder contains?
Thanks