How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?

by   Ferenc Huszár, et al.

Modern applications and progress in deep learning research have created renewed interest for generative models of text and of images. However, even today it is unclear what objective functions one should use to train and evaluate these models. In this paper we present two contributions. Firstly, we present a critique of scheduled sampling, a state-of-the-art training method that contributed to the winning entry to the MSCOCO image captioning benchmark in 2015. Here we show that despite this impressive empirical performance, the objective function underlying scheduled sampling is improper and leads to an inconsistent learning algorithm. Secondly, we revisit the problems that scheduled sampling was meant to address, and present an alternative interpretation. We argue that maximum likelihood is an inappropriate training objective when the end-goal is to generate natural-looking samples. We go on to derive an ideal objective function to use in this situation instead. We introduce a generalisation of adversarial training, and show how such method can interpolate between maximum likelihood training and our ideal training objective. To our knowledge this is the first theoretical analysis that explains why adversarial training tends to produce samples with higher perceived quality.


page 1

page 2

page 3

page 4


Semi-Implicit Generative Model

To combine explicit and implicit generative models, we introduce semi-im...

Coverage and Quality Driven Training of Generative Image Models

Generative modeling of natural images has been extensively studied in re...

Adversarial Robustness of Flow-Based Generative Models

Flow-based generative models leverage invertible generator functions to ...

Unsupervised Out-of-Distribution Detection with Batch Normalization

Likelihood from a generative model is a natural statistic for detecting ...

Sliced-Wasserstein normalizing flows: beyond maximum likelihood training

Despite their advantages, normalizing flows generally suffer from severa...

Improving image generative models with human interactions

GANs provide a framework for training generative models which mimic a da...

Maximum Likelihood Training of Autoencoders

Maximum likelihood training has favorable statistical properties and is ...

Please sign up or login with your details

Forgot password? Click here to reset