cr-gjx/ltf-label-transformation-framework

The codes of paper "LTF: A Label Transformation Framework for Correcting Target Shift"

ltf-label-transformation-framework's Issues

Questions about code

Hi authors
Interesting work! But I have a few questions about the code

1. The code seems a bit messy, (some unused defines)
   Also, only the 'f-m' implemention is correct?
2. In your zip file, your preTrain process .npy data seems not included
   Do I need to run the preTrain?
3. Could you release the continuous label shift codes?
4. confusing implementation:
   a) In your paper, you said LT takes in noise and source label, but in your code you just use random label?
   b) a bit confused about the reinforce loss, could you explain a little ?
   for example, why use this and what is 0.15 in the code

代码可以开源吗？

您好，代码可以分享下吗？《LTF: A Label Transformation Framework for Correcting Target Shift》