Quantification of Uncertainty with Adversarial Models

Kajetan Schweighofer, Lukas Aichberger, Mykyta Ielanskyi, Günter Klambauer, Sepp Hochreiter

Keywords: uncertainty, uncertainty quantification, predictive uncertainty, epistemic uncertainty, out of distribution, mc dropout, deep ensembles, sg-mcmc, adversarial model, adversarial model search, imagenet

TL;DR: We introduce QUAM to quantify predictive uncertainty with adversarial models (not adversarial examples!). Adversarial models identify important posterior modes that are missed by current uncertainty methods.

Abstract: Quantifying uncertainty is important for actionable predictions in real-world applications. A crucial part of predictive uncertainty quantification is the estimation of epistemic uncertainty, which is defined as an integral of the product between a divergence function and the posterior. Current methods such as Deep Ensembles or MC dropout underperform at estimating the epistemic uncertainty, since they primarily consider the posterior when sampling models. We suggest Quantification of Uncertainty with Adversarial Models (QUAM) to better estimate the epistemic uncertainty. QUAM identifies regions where the whole product under the integral is large, not just the posterior. Consequently, QUAM has lower approximation error of the epistemic uncertainty compared to previous methods. Models for which the product is large correspond to adversarial models (not adversarial examples!). Adversarial models have both a high posterior as well as a high divergence between their predictions and that of a reference model. Our experiments show that QUAM excels in capturing epistemic uncertainty for deep learning models and outperforms previous methods on challenging tasks in the vision domain.

Submission Number: 14046