code release for "Towards Discriminability and Diversity: Batch Nuclear-norm Maximization under Label Insufficient Situations" ( CVPR2020 oral)
We prove in the paper that Batch Nuclear-norm Maximization (BNM) could ensure the prediction discriminability and diversity, which is an effective method under label insufficient situations.
Assume X
is the prediction matrix. We could calculate BNM loss in both Pytorch and Tensorflow, as follows:
-Pytorch
- Direct calculation (Since there remains direct approach for nuclear-norm)
L_BNM = -torch.norm(X,'nuc')
- Calculation by SVD
L_BNM = -torch.sum(torch.svd(X, compute_uv=False)[1])
-Tensorflow
L_BNM = -tf.reduce_sum(tf.svd(X, compute_uv=False))
We apply BNM to domain adaptation (DA) in DA, unsupervised open domain recognition (UODR) in UODR and semi-supervised learning (SSL) in SSL.
Training instructions for DA, UODR and SSL are in the README.md
in DA, UODR and SSL respectively.
If you use this code for your research, please consider citing:
@article{cui2020nnm,
title={Towards Discriminability and Diversity: Batch Nuclear-norm Maximization under Label Insufficient Situations},
author={Cui, Shuhao and Wang, Shuhui and Zhuo, Junbao and Li, Liang and Huang, Qingming and Tian Qi},
journal={arXiv preprint arXiv:2003.12237},
year={2020}
}
If you have any problem about our code, feel free to contact
or describe your problem in Issues.