CoDi: Co-evolving Contrastive Diffusion Models for Mixed-type Tabular Synthesis

With growing attention to tabular data these days, the attempt to apply a synthetic table to various tasks has been expanded toward various scenarios. Owing to the recent advances in generative modeling, fake data generated by <PRE_TAG>tabular data synthesis</POST_TAG> models become sophisticated and realistic. However, there still exists a difficulty in modeling <PRE_TAG>discrete variables</POST_TAG> (columns) of tabular data. In this work, we propose to process continuous and discrete variables separately (but being conditioned on each other) by two diffusion models. The two <PRE_TAG>diffusion models</POST_TAG> are co-evolved during training by reading conditions from each other. In order to further bind the <PRE_TAG>diffusion models</POST_TAG>, moreover, we introduce a <PRE_TAG>contrastive learning</POST_TAG> method with a <PRE_TAG>negative sampling</POST_TAG> method. In our experiments with 11 real-world tabular datasets and 8 baseline methods, we prove the efficacy of the proposed method, called CoDi.