Out-of-Distribution Detection through Soft Clustering with Non-Negative Kernel Regression

TL;DR

This paper introduces a soft clustering method using non-negative kernel regression for out-of-distribution detection, significantly reducing computational and storage costs while outperforming existing methods on multiple benchmarks.

Contribution

The paper presents a novel soft clustering approach for OOD detection that is more efficient and effective than prior methods, including an entropy-constrained variant.

Findings

Up to 11x faster inference time

87% reduction in storage requirements

Outperforms existing methods by up to 4 AUROC points

Abstract

As language models become more general purpose, increased attention needs to be paid to detecting out-of-distribution (OOD) instances, i.e., those not belonging to any of the distributions seen during training. Existing methods for detecting OOD data are computationally complex and storage-intensive. We propose a novel soft clustering approach for OOD detection based on non-negative kernel regression. Our approach greatly reduces computational and space complexities (up to 11x improvement in inference time and 87% reduction in storage requirements) and outperforms existing approaches by up to 4 AUROC points on four different benchmarks. We also introduce an entropy-constrained version of our algorithm, which leads to further reductions in storage requirements (up to 97% lower than comparable approaches) while retaining competitive performance. Our soft clustering approach for OOD detection highlights its potential for detecting tail-end phenomena in extreme-scale data settings.