How to determine when the useful feature is learnt?

[Pixie8888]

Hi,

Thanks for sharing the code! I have a question about how to decide the useful feature is learnt. For example, when training on cifar-10, the first 130 epochs are only trained with given label without noise correction. How to decide the "130 epochs"?

[LanXiaoPang613]

Hi, i have some questions about the calculation of function "supervised_masks_estimation" and "unsupervised_masks_estimation", in which the mask2 is removed the diagonal, but actually after calculating the "quad1_sup" by the eq="quad1_sup = torch.eq(labels[mix_index1], labels.t()).float()" , the order of the matrix "quad1_sup"has been shuffled, so the diagonal is not going to be all ones. So I have a little confusion to this operation, looking forward to your replies.