DeepAI AI Chat
Log In Sign Up

Leveraging Recursive Gumbel-Max Trick for Approximate Inference in Combinatorial Spaces

by   Kirill Struminsky, et al.
Mail.Ru Group

Structured latent variables allow incorporating meaningful prior knowledge into deep learning models. However, learning with such variables remains challenging because of their discrete nature. Nowadays, the standard learning approach is to define a latent variable as a perturbed algorithm output and to use a differentiable surrogate for training. In general, the surrogate puts additional constraints on the model and inevitably leads to biased gradients. To alleviate these shortcomings, we extend the Gumbel-Max trick to define distributions over structured domains. We avoid the differentiable surrogates by leveraging the score function estimators for optimization. In particular, we highlight a family of recursive algorithms with a common feature we call stochastic invariant. The feature allows us to construct reliable gradient estimates and control variates without additional constraints on the model. In our experiments, we consider various structured latent variable models and achieve results competitive with relaxation-based counterparts.


page 1

page 2

page 3

page 4


Efficient Marginalization of Discrete and Structured Latent Variables via Sparsity

Training neural network models with discrete (categorical or structured)...

Low-variance Black-box Gradient Estimates for the Plackett-Luce Distribution

Learning models with discrete latent variables using stochastic gradient...

REBAR: Low-variance, unbiased gradient estimates for discrete latent variable models

Learning in models with discrete latent variables is challenging due to ...

Gradient Estimation with Stochastic Softmax Tricks

The Gumbel-Max trick is the basis of many relaxed gradient estimators. T...

Revisiting Reweighted Wake-Sleep

Discrete latent-variable models, while applicable in a variety of settin...

Stochastic polynomial chaos expansions to emulate stochastic simulators

In the context of uncertainty quantification, computational models are r...

Understanding the Mechanics of SPIGOT: Surrogate Gradients for Latent Structure Learning

Latent structure models are a powerful tool for modeling language data: ...