search

tim-learn / SHOT

code released for our ICML 2020 paper "Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation"
MIT License
437 stars 77 forks source link

Fair diversity Term #42

Open dodosly1 opened 7 months ago

dodosly1 commented 7 months ago

Hi :),

I am studying your UDA algorithm and I am trying to adapt it to my specific problem.

My question is the following:

  1. You seem to subtract the fair diversity loss from the entropy loss. Is this intentional? I've re-implemented your code from scratch and it seems that when i also do the same it I can reproduce your results

Initially i've implemented it using the KL divergence from the second equation of your paper and I wouldn't converge to your results.

Best Regards, Antonios Lykourinas