This repo contains code that supports the undergrad thesis that I wrote. The problem we tackle is data augmentation for soil microbiome samples. Among different generative models, a GAN is used. In particular, a Wasssertein GAN with gradient penalty.

This repo contains code for:

• Automating experiments
• Custom visualizations
• Metrics to assess the quality of samples. Usually, images would use an inception score, but our data is compositional.

This project borrowed some code for the WGAN-GP implementation from @ChengBinJin.