EBMs Trained with Maximum Likelihood are Generator Models Trained with a Self-adverserial Loss

02/23/2021
by   Zhisheng Xiao, et al.
5

Maximum likelihood estimation is widely used in training Energy-based models (EBMs). Training requires samples from an unnormalized distribution, which is usually intractable, and in practice, these are obtained by MCMC algorithms such as Langevin dynamics. However, since MCMC in high-dimensional space converges extremely slowly, the current understanding of maximum likelihood training, which assumes approximate samples from the model can be drawn, is problematic. In this paper, we try to understand this training procedure by replacing Langevin dynamics with deterministic solutions of the associated gradient descent ODE. Doing so allows us to study the density induced by the dynamics (if the dynamics are invertible), and connect with GANs by treating the dynamics as generator models, the initial values as latent variables and the loss as optimizing a critic defined by the very same energy that determines the generator through its gradient. Hence the term - self-adversarial loss. We show that reintroducing the noise in the dynamics does not lead to a qualitative change in the behavior, and merely reduces the quality of the generator. We thus show that EBM training is effectively a self-adversarial procedure rather than maximum likelihood estimation.

READ FULL TEXT

page 5

page 6

page 11

page 13

page 14

page 15

research
03/09/2021

Maximum Likelihood Estimation for Hawkes Processes with self-excitation or inhibition

In this paper, we present a maximum likelihood method for estimating the...
research
02/09/2016

Herding as a Learning System with Edge-of-Chaos Dynamics

Herding defines a deterministic dynamical system at the edge of chaos. I...
research
12/19/2014

On distinguishability criteria for estimating generative models

Two recently introduced criteria for estimation of generative models are...
research
01/04/2019

Coverage and Quality Driven Training of Generative Image Models

Generative modeling of natural images has been extensively studied in re...
research
07/11/2021

Dual Training of Energy-Based Models with Overparametrized Shallow Neural Networks

Energy-based models (EBMs) are generative models that are usually traine...
research
04/22/2019

On Learning Non-Convergent Short-Run MCMC Toward Energy-Based Model

This paper studies a curious phenomenon in learning energy-based model (...
research
06/04/2020

MLE-guided parameter search for task loss minimization in neural sequence modeling

Neural autoregressive sequence models are used to generate sequences in ...

Please sign up or login with your details

Forgot password? Click here to reset