JoyHuYY1412/DDE_CIL

Implementation of Distilling Causal Effect

lichong952012 opened this issue · 1 comments

Thanks for your excellent work. I have a question, in equation 6 of the article, how is the multiplication P (Y|I= i) *Wi calculated in the code? In loss1, the weight W is not included.

Thanks for your excellent work. I have a question, in equation 6 of the article, how is the multiplication P (Y|I= i) *Wi calculated in the code? In loss1, the weight W is not included.

@lichong952012 Hi, please check here

outputs_joint = (mu_1 * outputs_soft + mu_2 * outputs_match_soft) / (mu_1 + mu_2)

The logits are weighted summed.