ORTHOGONAL PERMUTATION SAMPLING FOR SHAPLEY VALUES: UNBIASED STRATIFIED ESTIMATORS WITH VARIANCE GUARANTEES

Authors

  • YASH VARSHNEY GURUKUL THE SCHOOL, INDIA https://orcid.org/0009-0007-7563-0567
  • RANAV TYAGI GURUKUL THE SCHOOL
  • ANURAG SINHA COMPUTER SCIENCE DEPARTMENT, ICFAI UNIVERSITY

DOI:

https://doi.org/10.64680/jisads.v3i2.44

Abstract

Shapley values for feature attribution suffer from high variance requiring thousands of model evaluations. We introduce Orthogonal Permutation Sampling (OPS), achieving provable variance reduction through: (i) exact position stratification, (ii) antithetic permutation coupling, and (iii) control variates. We prove finite-sample variance dominance over Monte Carlo and non-positive covariance under submodularity. Empirical validation across six benchmarks shows 5-26× variance reduction for typical dimensions (n=10-20) and 67× for n=50. OPS achieves 2-5× lower MSE than KernelSHAP at equivalent budgets with 7% runtime overhead (all p<0.001). The framework is model-agnostic, maintains exact unbiasedness, scales linearly to n=100, and provides production-ready reliable feature attributions.

Downloads

Published

2026-01-11

Issue

Vol. 3 No. 2 (2025): Journal of Intelligent Systems and applied data science (JISADS)

Section

Research and Review articles

License

Copyright (c) 2026 YASH VARSHNEY, RANAV TYAGI, ANURAG SINHA

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

You are free to:

  1. Share — copy and redistribute the material in any medium or format for any purpose, even commercially.
  2. Adapt — remix, transform, and build upon the material for any purpose, even commercially.
  3. The licensor cannot revoke these freedoms as long as you follow the license terms.

Under the following terms:

  1. Attribution — You must give appropriate credit , provide a link to the license, and indicate if changes were made . You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
  2. No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Notices:

You do not have to comply with the license for elements of the material in the public domain or where your use is permitted by an applicable exception or limitation .

No warranties are given. The license may not give you all of the permissions necessary for your intended use. For example, other rights such as publicity, privacy, or moral rights may limit how you use the material.