mikigom / DNPL-PyTorch

Official Code for "On the Power of Deep but Naive Partial Label Learning" (ICASSP 21)

Deep Naive Partial Label Learning

This repo offers official implementation for On the Power of Deep but Naive Partial Label Learning (ICASSP 21).

Reproduction

For reproducing results for Yahoo! dataset,

python dnpl.py -dset yahoo -model medium

Datasets

All datasets used in this work are provided by Min-Ling ZHANG.

Lost: http://palm.seu.edu.cn/zhangml/files/lost.rar
MSRCv2: http://palm.seu.edu.cn/zhangml/files/MSRCv2.rar
Soccer: http://palm.seu.edu.cn/zhangml/files/Soccer%20Player.rar
Yahoo! News: http://palm.seu.edu.cn/zhangml/files/Yahoo!%20News.rar

Reference

@inproceedings{seo2021power,
  title={On the Power of Deep but Naive Partial Label Learning},
  author={Seo, Junghoon and Huh, Joon Suk},
  journal={IEEE International Conference on Acoustics, Speech, and Signal Processing},
  year={2021}
}