A Simple and Effective Method for Uncertainty Quantification and OOD Detection
This work addresses uncertainty quantification and OOD detection for machine learning practitioners, offering a more efficient alternative to existing methods, though it appears incremental as it builds on kernel density estimation techniques.
The paper tackles the computational and storage inefficiencies of Bayesian neural networks and deep ensembles for uncertainty quantification by proposing a single deterministic model that uses feature space density for distributional shift detection and OOD detection, showing improved performance on synthetic datasets and CIFAR-10 vs. SVHN tasks.
Bayesian neural networks and deep ensemble methods have been proposed for uncertainty quantification; however, they are computationally intensive and require large storage. By utilizing a single deterministic model, we can solve the above issue. We propose an effective method based on feature space density to quantify uncertainty for distributional shifts and out-of-distribution (OOD) detection. Specifically, we leverage the information potential field derived from kernel density estimation to approximate the feature space density of the training set. By comparing this density with the feature space representation of test samples, we can effectively determine whether a distributional shift has occurred. Experiments were conducted on a 2D synthetic dataset (Two Moons and Three Spirals) as well as an OOD detection task (CIFAR-10 vs. SVHN). The results demonstrate that our method outperforms baseline models.