Long-Tailed Out-of-Distribution Detection via Normalized Outlier Distribution Adaptation

TL;DR

To obtain a more reliable set of predicted OOD samples on long-tailed ID data, a novel dual-normalized energy loss is introduced in AdaptOD, which leverages class- and sample-wise normalized energy to enforce a more balanced prediction energy on imbalanced ID samples.

Abstract

One key challenge in Out-of-Distribution (OOD) detection is the absence of ground-truth OOD samples during training. One principled approach to address this issue is to use samples from external datasets as outliers (i.e., pseudo OOD samples) to train OOD detectors. However, we find empirically that the outlier samples often present a distribution shift compared to the true OOD samples, especially in Long-Tailed Recognition (LTR) scenarios, where ID classes are heavily imbalanced, \ie, the true OOD samples exhibit very different probability distribution to the head and tailed ID classes from the outliers. In this work, we propose a novel approach, namely normalized outlier distribution adaptation (AdaptOD), to tackle this distribution shift problem. One of its key components is dynamic outlier distribution adaptation that effectively adapts a vanilla outlier distribution based on the outlier samples to the true OOD distribution by utilizing the OOD knowledge in the predicted OOD samples during inference. Further, to obtain a more reliable set of predicted OOD samples on long-tailed ID data, a novel dual-normalized energy loss is introduced in AdaptOD, which leverages class- and sample-wise normalized energy to enforce a more balanced prediction energy on imbalanced ID samples. This helps avoid bias toward the head samples and learn a substantially better vanilla outlier distribution than existing energy losses during training. It also eliminates the need of manually tuning the sensitive margin hyperparameters in energy losses. Empirical results on three popular benchmarks for OOD detection in LTR show the superior performance of AdaptOD over state-of-the-art methods. Code is available at https://github.com/mala-lab/AdaptOD.

Figures (4)

• Figure 1: (a) Mean energy distribution on six OOD datasets with CIFAR100-LT cao2019learning as ID data. (b) The results of the energy distribution of the OOD dataset SVHN netzer2011reading using our proposed dynamic outlier distribution adaptation (DODA) and an existing energy loss EnergyOE liu2020energy, where the outlier data is taken from TinyImages80M torralba200880. (c) The results of using both of our proposed DODA and dual-normalized energy loss (DNE). (d) The ratio of the energy of each ID class to the aggregated energy of all ID classes.
• Figure 2: Overview of AdaptOD, which consists of two components, DODA (Left) and DNE (Right). Left: Each test sample is assigned a global energy-based OOD score $\mathbb G(x)$ to adapt the outlier distribution $\mathcal{P}^{out}$. DODA then uses the adapted outlier distribution $\mathcal{P}^{out}$ to calibrate the global energy score $\mathbb G(x)$, obtaining the calibrated global energy score $\mathbb G^{\mathcal{P}}(x)$ as the OOD score. Right: For each iteration, DNE first applies Batch Energy Normalization on logit output to obtain the normalized energy, and then utilizes this energy to optimize a dual energy loss function at both the class and sample levels.
• Figure 3: The average performance over six OOD datasets on CIFAR100-LT with an increasing percentage of true OOD samples fed to TTA methods.
• Figure 4: Average performance of AdaptOD w.r.t. $\alpha$ over six OOD datasets on CIFAR100-LT.