Updates on the Complexity of SHAP Scores

Authors: Xuanxiang Huang, Joao Marques-Silva

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical This paper refines some of the existing complexity claims, including families of classifiers for which the computation of SHAP scores is computationally hard and those for which there exist polynomial-time algorithms. Moreover, this paper proves that the computational complexity of SHAP scores is #P-hard for these widely used families of random forest (RF) classifiers.
Researcher Affiliation Academia Xuanxiang Huang1 , Joao Marques-Silva2 1CNRS@CREATE, Singapore 2ICREA, University of Lleida, Spain
Pseudocode Yes Algorithm 1 Computation of SHAP scores for non-boolean DTs (Page 8)
Open Source Code No The paper does not provide any concrete access to source code for the methodology described.
Open Datasets No The paper is theoretical and does not involve training models on datasets.
Dataset Splits No The paper is theoretical and does not involve dataset splits for validation.
Hardware Specification No The paper focuses on theoretical analysis and proofs, and does not describe experiments that would require hardware specifications.
Software Dependencies No The paper focuses on theoretical analysis and proofs, and does not describe experiments that would require software dependencies.
Experiment Setup No The paper focuses on theoretical analysis and proofs, and does not describe an experimental setup.