Number of transformation in subpolicy

[kakusikun]

Thanks for your work on pytorch.

In the searching space, you find two of extra transformations here.

Why two? It's better to find the combination of arbitrary transformations, isn't it?

Finally, only one transformation is picked randomly here, why??

If there are any misunderstanding, please correct me, thx.

[kakusikun]

Sorry, after I read the paper, the two operation in a subpolicy is the idea from AutoAugment.

[kakusikun]

The top10 subpolicies will be in here My final question should be "why don't you merge the subpolicies into one but randomly choose?"

[JunYeopLee]

Hi @allenfutaki . Thank you for your question.

''My final question should be "why don't you merge the sub-policies into one but randomly choose?"'' >> I think that each sub-policy has to be applied in an independent manner, not a sequential manner. Is that right? So I apply randomly chosen sub-policy per every training step.