Reproduce “VAEM: a Deep Generative Model for Heterogeneous Mixed Type Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chao Ma, Sebastian Tschiatschek, José Miguel Hernández-Lobato, Richard Turner, Cheng Zhang
VAEM models heterogeneous tables for generation, conditional prediction, and sequential feature selection. It first fits a separate, type-appropriate variational autoencoder for each column, then learns relationships among their latent representations with a second model. On five datasets, it achieves the best average generative-modeling rank, 1.40 ± 0.36, though vanilla VAE performs better on Energy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.