Uncertainty Quantification via Stable Distribution Propagation
Stanford University · SLAC National Laboratory · Rheinische Friedrich-Wilhelms-Universität Bonn, Rheinische Friedrich-Wilhelms Universität Bonn · Paris-Lodron-University of Salzburg · University of Konstanz · IBM Research, MIT-IBM Watson AI Lab
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
We propose a new approach for propagating stable probability distributions through neural networks. Our method is based on local linearization, which we show to be an optimal approximation in terms of total variation distance for the ReLU non-linearity. This allows propagating Gaussian and Cauchy input uncertainties through neural networks to quantify their output uncertainties. To demonstrate the utility of propagating distributions, we apply the proposed method to predicting calibrated confidence intervals and selective prediction on out-of-distribution data. The results demonstrate a broad applicability of propagating distributions and show the advantages of our method over other approaches such as moment matching.