Here are the codes for our submitted paper. We modified the contrastive loss to the min-max loss where positive samples and negative samples are discriminated by series distance. The min-max loss works better than the contrastive loss.

https://github.com/zhkai/IGCL/releases/tag/dataset. And you can find and download all datasets at https://github.com/decisionintelligence/TAB (Data preparation).

• Python 3.6
• tensorflow 1.14.0
• pytorch 1.4.0

run with 'python main.py'