Learning Multi-Manifold Embedding for Out-Of-Distribution Detection

TL;DR

This paper proposes a Multi-Manifold Embedding Learning framework that enhances out-of-distribution detection by jointly optimizing hypersphere and hyperbolic spaces, achieving high accuracy with minimal OOD samples and no retraining.

Contribution

It introduces a novel multi-manifold embedding approach for OOD detection that outperforms existing methods and requires fewer OOD samples without retraining.

Findings

Significant reduction in false positive rate (FPR) compared to state-of-the-art methods.

High AUC maintained with minimal OOD samples and no retraining.

Comparable performance to methods using millions of outlier samples.

Abstract

Detecting out-of-distribution (OOD) samples is crucial for trustworthy AI in real-world applications. Leveraging recent advances in representation learning and latent embeddings, Various scoring algorithms estimate distributions beyond the training data. However, a single embedding space falls short in characterizing in-distribution data and defending against diverse OOD conditions. This paper introduces a novel Multi-Manifold Embedding Learning (MMEL) framework, optimizing hypersphere and hyperbolic spaces jointly for enhanced OOD detection. MMEL generates representative embeddings and employs a prototype-aware scoring function to differentiate OOD samples. It operates with very few OOD samples and requires no model retraining. Experiments on six open datasets demonstrate MMEL's significant reduction in FPR while maintaining a high AUC compared to state-of-the-art distance-based OOD detection methods. We analyze the effects of learning multiple manifolds and visualize OOD score distributions across datasets. Notably, enrolling ten OOD samples without retraining achieves comparable FPR and AUC to modern outlier exposure methods using 80 million outlier samples for model training.