neural_kernels_code icon indicating copy to clipboard operation
neural_kernels_code copied to clipboard

public dataset

Open Kid-key opened this issue 4 years ago • 3 comments

Hi, I can't access AWS services currently, would you provide the dataset via another approach like google drive?

Kid-key avatar Jun 19 '21 04:06 Kid-key

Which dataset would you like to access? I can create a public url for you.

Vaishaal avatar Jul 19 '21 16:07 Vaishaal

I has modified several code and tried on vanilla cifar-10 dataset (http://www.cs.toronto.edu/~kriz/cifar.html) with zero-mean and normalization pre-processing, however, my accuracy for Myrtle5 is only about 76%. Such, I want to try your cifar-10 dataset as my baseline, thanks.

Kid-key avatar Jul 20 '21 13:07 Kid-key

Ah you need to use ZCA pre-processing as mentioned in the paper. 76 seems correct for standard pre_processing. Here is a link for ZCA pre-processed cifar-10: https://imagenetv2public.s3.us-west-2.amazonaws.com/cifar_zca.npz

Vaishaal avatar Sep 27 '21 07:09 Vaishaal