Direct Optimization through $\arg \max$ for Discrete Variational Auto-Encoder

Guy Lorberbom, Tamir Hazan

Abstract: Reparameterization of variational auto-encoders is an effective method for reducing the variance of their gradient estimates. However, when the latent variables are discrete, a reparameterization is problematic due to discontinuities in the discrete space. In this work, we extend the direct loss minimization technique to discrete variational auto-encoders. We first reparameterize a discrete random variable using the $\arg \max$ function of the Gumbel-Max perturbation model. We then use direct optimization to propagate gradients through the non-differentiable $\arg \max$ using two perturbed $\arg \max$ operations.

Keywords: discrete variational auto encoders, generative models, perturbation models

Code: [ GuyLor/direct_vae](https://github.com/GuyLor/direct_vae) + [ 1 community implementation](https://paperswithcode.com/paper/?openreview=S1ey2sRcYQ)

Data: [CelebA](https://paperswithcode.com/dataset/celeba), [Fashion-MNIST](https://paperswithcode.com/dataset/fashion-mnist)

Community Implementations: [ 4 code implementations](https://www.catalyzex.com/paper/direct-optimization-through-arg-max-for/code)