Using Perturbation to Improve Goodness-of-Fit Tests based on Kernelized Stein Discrepancy

Xing Liu, Andrew B. Duncan, Axel Gandy
4/28/2023
stat.MLcs.LGstat.ME

No official code found. Best implementation: XingLLiu/pksd (3 stars, MIT, last commit 2024-06-14).

Abstract

Kernelized Stein discrepancy (KSD) is a score-based discrepancy widely used in goodness-of-fit tests. It can be applied even when the target distribution has an unknown normalising factor, such as in Bayesian analysis. We show theoretically and empirically that the KSD test can suffer from low power when the target and the alternative distributions have the same well-separated modes but differ in mixing proportions. We propose to perturb the observed sample via Markov transition kernels, with respect to which the target distribution is invariant. This allows us to then employ the KSD test on the perturbed sample. We provide numerical evidence that with suitably chosen transition kernels the proposed approach can lead to substantially higher power than the KSD test.

View on arXivView PDF

Code Implementations(1)

XingLLiu/pksdCommunity100%
MIT

Code for the paper "Using Perturbation to Improve Goodness-of-Fit Tests based on Kernelized Stein Discrepancy".

30R, ShellApr 30, 20232 years agoMIT

Similar Papers

On the Robustness of Kernel Goodness-of-Fit Tests

85% match1 repo22024

Nyström Kernel Stein Discrepancy

84% match1 repo22024

Fast and Scalable Score-Based Kernel Calibration Tests

83% match2025

Correcting Mode Proportion Bias in Generalized Bayesian Inference via a Weighted Kernel Stein Discrepancy

82% match2025

The Polynomial Stein Discrepancy for Assessing Moment Convergence

82% match2024

Semiparametric KSD test: unifying score and distance-based approaches for goodness-of-fit testing

81% match2025

Cite this paper

@article{liu2023using,
  title  = {Using Perturbation to Improve Goodness-of-Fit Tests based on Kernelized Stein Discrepancy},
  author = {Xing Liu and Andrew B. Duncan and Axel Gandy},
  year   = {2023},
  eprint = {2304.14762},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2304.14762v3},
  journal = {ICML 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2304.14762.svg)](https://codeofpaper.com/paper/2304.14762)

Discussion