Nearest-neighbor class changes and entropy across network layers provide a post-hoc uncertainty score that outperforms softmax confidence on CIFAR-10 and partly on MNIST.
Uncertainty Estimation Using a Single Deep Deterministic Neural Network
1 Pith paper cite this work, alongside 233 external citations. Polarity classification is still indexing.
1
Pith paper citing it
233
external citations · Pith
abstract
We propose a method for training a deterministic deep model that can find and reject out of distribution data points at test time with a single forward pass. Our approach, deterministic uncertainty quantification (DUQ), builds upon ideas of RBF networks. We scale training in these with a novel loss function and centroid updating scheme and match the accuracy of softmax models. By enforcing detectability of changes in the input using a gradient penalty, we are able to reliably detect out of distribution data. Our uncertainty quantification scales well to large datasets, and using a single model, we improve upon or match Deep Ensembles in out of distribution detection on notable difficult dataset pairs such as FashionMNIST vs. MNIST, and CIFAR-10 vs. SVHN.
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Framework for Uncertainty Quantification Based on Nearest Neighbors Across Layers
Nearest-neighbor class changes and entropy across network layers provide a post-hoc uncertainty score that outperforms softmax confidence on CIFAR-10 and partly on MNIST.