Skip to content

Latest commit

 

History

History
30 lines (21 loc) · 995 Bytes

README.md

File metadata and controls

30 lines (21 loc) · 995 Bytes

Dynamic Importance Weighting (DIW)

This is a reproducing code for Dynamic Importance Weighting (DIW) in the NeurIPS'20 paper: Rethinking Importance Weighting for Deep Learning under Distribution Shift.

Link to the paper: https://proceedings.neurips.cc//paper/2020/file/8b9e7ab295e87570551db122a04c6f7c-Paper.pdf

Requirements

The code was developed and tested based on the following environment.

  • python 3.8
  • pytorch 1.6.0
  • torchvision 0.7.0
  • cudatoolkit 10.2
  • cvxopt 1.2.0
  • matplotlib
  • sklearn
  • tqdm

Quick start

You can run an example code of DIW on Fashion-MNIST under 0.4 symmetric label noise.

python diw.py

Example result

After running python diw.py, a output figure and text file of test accurary are made in ./output/ by default.

Citation

If the code is useful in your research, please cite the following:
Tongtong Fang, Nan Lu, Gang Niu, Masashi Sugiyama. Rethinking Importance Weighting for Deep Learning under Distribution Shift. NeurIPS 2020.