Odd-One-Out Representation Learning

TL;DR

This paper introduces an odd-one-out based weakly-supervised evaluation method for representation learning, demonstrating its effectiveness in model selection and outperforming existing disentanglement models on visual reasoning tasks.

Contribution

It proposes a novel weakly-supervised evaluation approach using odd-one-out observations and a metric-learning VAE that excels in abstract visual reasoning tasks.

Findings

High correlation between odd-one-out task performance and downstream task success

The proposed VAE outperforms standard unsupervised and weakly-supervised models

Effective model selection method for representation learning in complex tasks

Abstract

The effective application of representation learning to real-world problems requires both techniques for learning useful representations, and also robust ways to evaluate properties of representations. Recent work in disentangled representation learning has shown that unsupervised representation learning approaches rely on fully supervised disentanglement metrics, which assume access to labels for ground-truth factors of variation. In many real-world cases ground-truth factors are expensive to collect, or difficult to model, such as for perception. Here we empirically show that a weakly-supervised downstream task based on odd-one-out observations is suitable for model selection by observing high correlation on a difficult downstream abstract visual reasoning task. We also show that a bespoke metric-learning VAE model which performs highly on this task also out-performs other standard unsupervised and a weakly-supervised disentanglement model across several metrics.