Meta Learning Low Rank Covariance Factors for Energy-Based Deterministic Uncertainty

TL;DR

This paper introduces a meta-learning approach to efficiently construct task-specific covariance matrices for uncertainty estimation using low-rank factors, improving calibration under dataset shift.

Contribution

It proposes a novel meta-learning method for low-rank covariance factors with an attentive set encoder, enhancing uncertainty estimation across tasks.

Findings

Meta-learning improves covariance matrix estimation.

Low-rank factors enhance uncertainty calibration.

Method performs well under dataset shift.

Abstract

Numerous recent works utilize bi-Lipschitz regularization of neural network layers to preserve relative distances between data instances in the feature spaces of each layer. This distance sensitivity with respect to the data aids in tasks such as uncertainty calibration and out-of-distribution (OOD) detection. In previous works, features extracted with a distance sensitive model are used to construct feature covariance matrices which are used in deterministic uncertainty estimation or OOD detection. However, in cases where there is a distribution over tasks, these methods result in covariances which are sub-optimal, as they may not leverage all of the meta information which can be shared among tasks. With the use of an attentive set encoder, we propose to meta learn either diagonal or diagonal plus low-rank factors to efficiently construct task specific covariance matrices. Additionally, we propose an inference procedure which utilizes scaled energy to achieve a final predictive distribution which is well calibrated under a distributional dataset shift.