Eigenvalue-based OOD detection framework.
Out-of-distribution (OOD) detection is critical for the safe deployment of machine learning systems in safety-sensitive domains. Diffusion models have recently emerged as powerful generative models, capable of capturing complex data dis- tributions through iterative denoising. Building on this progress, recent work has explored their potential for OOD detection. We propose EigenScore, a new OOD detection method that leverages the eigenvalue spectrum of the posterior covariance induced by a diffusion model. We argue that posterior covariance provides a con- sistent signal of distribution shift, leading to larger trace and leading eigenvalues on OOD inputs, yielding a clear spectral signature. We further provide analysis explicitly linking posterior covariance to distribution mismatch, establishing it as a reliable signal for OOD detection. To ensure tractability, we adopt a Jacobian-free subspace iteration method to estimate the leading eigenvalues using only forward evaluations of the denoiser. Empirically, EigenScore achieves state-of-the-art per- formance, with up to 5% AUROC improvement over the best baseline. Notably, it remains robust in near-OOD settings such as CIFAR-10 vs CIFAR-100, where existing diffusion-based methods often fail.
Download datasets CelebA, Cifar10, Cifar100, SVHN, and TinyImages. Download Models
Create the conda virtual enviroment.
conda env create -f EigenScore.yml
conda activate eigenscoreRead the description within the file and adjust the corresponding arguments.
python eigenscore_launcher.py
Usage Example
python eigenscore_ood_detection.py -id <id dataset> -ood <ood-dataset> -e <eigenvalues file directory> -time <timesteps> -a <aggregation method> -rep <number of repetitions> -t <test fraction>Read the description within the file and adjust the corresponding arguments.
python ES_optimize_parameters.pyRead the description within the file and adjust the corresponding arguments.
python ES_return_best_parameters.py