Out-of-distribution Detection via Frequency-regularized Generative Models
This addresses a critical safety issue for AI models in open-world deployments by improving OOD detection, though it is an incremental advance over existing methods.
The paper tackles the problem of deep generative models assigning high likelihood to out-of-distribution (OOD) inputs by proposing a frequency-regularized learning (FRL) framework that incorporates high-frequency information to focus on semantically relevant features, achieving state-of-the-art performance with a 10.7% improvement in AUROC and 147x faster inference speed compared to a baseline.
Modern deep generative models can assign high likelihood to inputs drawn from outside the training distribution, posing threats to models in open-world deployments. While much research attention has been placed on defining new test-time measures of OOD uncertainty, these methods do not fundamentally change how deep generative models are regularized and optimized in training. In particular, generative models are shown to overly rely on the background information to estimate the likelihood. To address the issue, we propose a novel frequency-regularized learning FRL framework for OOD detection, which incorporates high-frequency information into training and guides the model to focus on semantically relevant features. FRL effectively improves performance on a wide range of generative architectures, including variational auto-encoder, GLOW, and PixelCNN++. On a new large-scale evaluation task, FRL achieves the state-of-the-art performance, outperforming a strong baseline Likelihood Regret by 10.7% (AUROC) while achieving 147$\times$ faster inference speed. Extensive ablations show that FRL improves the OOD detection performance while preserving the image generation quality. Code is available at https://github.com/mu-cai/FRL.