Inverting Variational Autoencoders for Improved Generative Accuracy
This work addresses the challenge of leveraging untapped data sources in semi-supervised learning for domains like spectroscopy and digit recognition, though it appears incremental as it builds on existing variational autoencoder methods.
The paper tackles the problem of semi-supervised learning with deep generative models by exploiting unlabeled data in structured codomains, resulting in improved disentanglement of latent variables and better discriminative prediction on Martian spectroscopic and handwritten digit datasets.
Recent advances in semi-supervised learning with deep generative models have shown promise in generalizing from small labeled datasets ($\mathbf{x},\mathbf{y}$) to large unlabeled ones ($\mathbf{x}$). In the case where the codomain has known structure, a large unfeatured dataset ($\mathbf{y}$) is potentially available. We develop a parameter-efficient, deep semi-supervised generative model for the purpose of exploiting this untapped data source. Empirical results show improved performance in disentangling latent variable semantics as well as improved discriminative prediction on Martian spectroscopic and handwritten digit domains.