search

melodyguan / enas

TensorFlow Code for paper "Efficient Neural Architecture Search via Parameter Sharing"
https://arxiv.org/abs/1802.03268
Apache License 2.0
1.58k stars 390 forks source link

drop path explanation #37

Open bkj opened 6 years ago

bkj commented 6 years ago

Hi --

Thanks again for releasing the code. Are you able to explain a little more what _apply_drop_path is doing in cifar10/micro_child.py, and what the motivation for this regularization is?

Thanks Ben

karandwivedi42 commented 6 years ago

You can see this paper (Sec 3.1 specifically)

ashaw596 commented 6 years ago

Hmm. I was looking through the code. So is that only used in the final training? It looks like its only called in _fixed_layer.