HAct: Out-of-Distribution Detection with Neural Net Activation Histograms
This addresses the problem of detecting anomalous data for deployed neural networks, offering a simple and efficient solution with strong performance gains.
The paper tackles out-of-distribution detection for neural networks by proposing HAct, a method using activation histograms, and achieves a true positive rate of 95% with only 0.03% false positives on standard benchmarks, outperforming previous state-of-the-art by 20.67% in false positive rate.
We propose a simple, efficient, and accurate method for detecting out-of-distribution (OOD) data for trained neural networks. We propose a novel descriptor, HAct - activation histograms, for OOD detection, that is, probability distributions (approximated by histograms) of output values of neural network layers under the influence of incoming data. We formulate an OOD detector based on HAct descriptors. We demonstrate that HAct is significantly more accurate than state-of-the-art in OOD detection on multiple image classification benchmarks. For instance, our approach achieves a true positive rate (TPR) of 95% with only 0.03% false-positives using Resnet-50 on standard OOD benchmarks, outperforming previous state-of-the-art by 20.67% in the false positive rate (at the same TPR of 95%). The computational efficiency and the ease of implementation makes HAct suitable for online implementation in monitoring deployed neural networks in practice at scale.