Estimating the Success of Unsupervised Image to Image Translation

TL;DR

This paper introduces a new theoretical bound to predict the success of unsupervised image-to-image translation, aiding hyperparameter tuning and stopping criteria without relying on labeled data.

Contribution

It proposes a novel bound based on the Simplicity Principle for unsupervised cross-domain mapping, applicable both in expectation and per sample.

Findings

The bound effectively predicts translation success across multiple algorithms.

It assists in hyperparameter selection and stopping point determination.

Experimental results validate the bound's practical utility.

Abstract

While in supervised learning, the validation error is an unbiased estimator of the generalization (test) error and complexity-based generalization bounds are abundant, no such bounds exist for learning a mapping in an unsupervised way. As a result, when training GANs and specifically when using GANs for learning to map between domains in a completely unsupervised way, one is forced to select the hyperparameters and the stopping epoch by subjectively examining multiple options. We propose a novel bound for predicting the success of unsupervised cross domain mapping methods, which is motivated by the recently proposed Simplicity Principle. The bound can be applied both in expectation, for comparing hyperparameters and for selecting a stopping criterion, or per sample, in order to predict the success of a specific cross-domain translation. The utility of the bound is demonstrated in an extensive set of experiments employing multiple recent algorithms. Our code is available at https://github.com/sagiebenaim/gan_bound .