Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

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

About

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).

Jie Ren, Stanislav Fort, Jeremiah Liu, Abhijit Guha Roy, Shreyas Padhy, Balaji Lakshminarayanan• 2021

Related benchmarks

TaskDatasetResultRank
Image ClassificationImageNet-1k (val)--
1469
Out-of-Distribution DetectioniNaturalist
AUROC96.1
219
Out-of-Distribution DetectionTextures
AUROC0.8938
168
Out-of-Distribution DetectionOpenOOD average of NINCO, iNat, SSB-Hard, OpenImages-O, Textures
FPR @ 95%26.9
130
Out-of-Distribution DetectionImageNet
FPR9533.1
108
Out-of-Distribution DetectionOpenImage-O
AUROC92.32
107
Near-OOD DetectionCIFAR-100 Near-OOD (test)
AUROC88.83
93
OOD DetectionCIFAR-10
FPR@9525.35
85
Out-of-Distribution DetectionImageNet-1k Textures ID OOD
AUROC89.38
85
Out-of-Distribution DetectionNINCO
AUROC87.31
82
Showing 10 of 83 rows
...

Other info

Follow for update