My contributions to COT-GAN paper (15 June submitted to arXiv) prior to NeurIPS review:
* designed all experiments (datasets, baselines and performance metrics);
* discovered the bias issue in Section 3.3, contributed to disucssion on mixed-Sinkhorn divergence (eqn 3.7), produced Figure 1 and proved Remark A.3;
* carried out all experiments in Section 5.1 and animated action dataset in Section 5.2, produced Figures 2, 3 and 5-8;
* substantial paper writing.