Understanding d-sne and application

[lchunleo]

Hi

Thanks for sharing your work.

Based on your code, it seem one network is used to be trained (LeNet) whereas in the paper, the source and target has their separate networks , where the learning takes place to max-min the loss based on their generated features. I may have misread your code or the paper. Could you advise me?