LGJun 16, 2021

A Simple Fix to Mahalanobis Distance for Improving Near-OOD Detection

arXiv:2106.09022v1297 citations
Originality Incremental advance
AI Analysis

This work addresses near-OOD detection for neural networks in vision, language, and biology domains, offering an incremental improvement over existing methods.

The paper tackled the problem of near out-of-distribution (OOD) detection by analyzing failure modes of Mahalanobis distance and proposing a simple fix called relative Mahalanobis distance, which improved performance by up to 15% AUROC on genomics OOD benchmarks.

Mahalanobis distance (MD) is a simple and popular post-processing method for detecting out-of-distribution (OOD) inputs in neural networks. We analyze its failure modes for near-OOD detection and propose a simple fix called relative Mahalanobis distance (RMD) which improves performance and is more robust to hyperparameter choice. On a wide selection of challenging vision, language, and biology OOD benchmarks (CIFAR-100 vs CIFAR-10, CLINC OOD intent detection, Genomics OOD), we show that RMD meaningfully improves upon MD performance (by up to 15% AUROC on genomics OOD).

Code Implementations4 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes