CUED Publications database

Getting a CLUE: A Method for Explaining Uncertainty Estimates

Antorán, J and Bhatt, U and Adel, T and Weller, A and Hernández-Lobato, JM Getting a CLUE: A Method for Explaining Uncertainty Estimates. In: ICLR 2021: Ninth International Conference on Learning Representations, 2021-5-3 to 2021-5-7. (Unpublished)

Full text not available from this repository.

Abstract

Both uncertainty estimation and interpretability are important factors for trustworthy machine learning systems. However, there is little work at the intersection of these two areas. We address this gap by proposing a novel method for interpreting uncertainty estimates from differentiable probabilistic models, like Bayesian Neural Networks (BNNs). Our method, Counterfactual Latent Uncertainty Explanations (CLUE), indicates how to change an input, while keeping it on the data manifold, such that a BNN becomes more confident about the input's prediction. We validate CLUE through 1) a novel framework for evaluating counterfactual explanations of uncertainty, 2) a series of ablation experiments, and 3) a user study. Our experiments show that CLUE outperforms baselines and enables practitioners to better understand which input patterns are responsible for predictive uncertainty.

Item Type: Conference or Workshop Item (UNSPECIFIED)
Uncontrolled Keywords: stat.ML stat.ML cs.LG
Subjects: UNSPECIFIED
Divisions: Div F > Computational and Biological Learning
Depositing User: Cron Job
Date Deposited: 26 Jun 2020 20:01
Last Modified: 27 May 2021 03:34
DOI: