Supplementary Material for One Loss for All: Deep Hashing with a Single Cosine Similarity based Learning Objective

Neural Information Processing Systems 

In Appendix B, we explain in details the proof in the main paper.