Question about the implementation of SiMix

[Noahsark]

Hi @yash0307,

Thanks for sharing the code and I feel its an interesting paper. My concern is that I didn't find anywhere you implement the method of SiMix which is described in paper. For my understanding, you are try to calculate the sims for mixup features without physically mixup them. But your code seems calculate all sims for both original and mixup samples. May I ask if this is my misunderstanding or you haven't released the code for SiMix yet? When will you release them? Thank you so much.

Best Regards.