Created by Shaohui Liu, Xiao Zhang, Jianqiao Wangni, and Jianbo Shi. This repo includes the source code implementation for "Normalized Diversification" in both PyTorch and TensorFlow.

The Normalized Diversification loss, computed over normalized pairwise distance matrices of both the latent vectors and the corresponding outputs in a batchwise manner, was initially described in arXiv to ameliorate the mode collapse problem. We release code on synthetic examples and image generation in this repo. Please follow the instructions in each folder.

@inproceedings{liu2019normalized,
  title={Normalized Diversification},
  author={Liu, Shaohui and Zhang, Xiao and Wangni, Jianqiao and Shi, Jianbo},
  booktitle={Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2019}
}