Sirius083

Thanks for replying, I am using the exact same data augmentation as your code(one floating point channel mean and channel standard deviation, 4-pixel padding, flipping). I am further compare the...

> Did you use data augmentation? (Padding, cropping and flipping on input image). If you didn't, it could overfit on training set and the test loss could increase. Thanks for...

Any update? Thanks a lot @Lyken17

My implementation of sparsenet-BC on cifar100 (d=100, k=32,64,128) validation error is 20.24%, which is 2% percent higher than 18.22%, I think it is due to the layer_per_stage(parameter in my implementaion...

For other network settings, I get similiary outputs in paper For example, model-BC d_100_k_24 is 22.71%(your), 22.97%(my) @Lyken17

Thanks, can you tell me the standard deviation of cifar100 under your implementation? I am using tensorflow to try you code. Thanks

https://gist.github.com/weiaicunzai/e623931921efefd4c331622c344d8151 Sorry to bother, I find the answer here Thanks

In em-routing's m-step, why not updating mean in the first two iterations? capsule_em.py L352? (the else part does not clear), can you give some explaination here, thanks