Home / Curated / Algorithms / Kernel SHAP

Kernel SHAP

Algorithms AIMech:KernelSHAP DRAFT

A model-agnostic Shapley-attribution estimator using weighted linear regression on feature-presence subsets.

Identity Notes & Concept Scope Kernel SHAP is a specific estimator within the SHAP framework. Its Shapley kernel and constraints distinguish it from an arbitrary LIME regression surrogate.

Claims & Grounded Evidence

1 Claim

Mechanism #explanation-method

Fit an additive explanation model to subset-evaluation outputs with the Shapley weighting kernel and constraints fixing the empty-subset baseline and full-input prediction; sampled subsets give an approximate estimate.

supports Source: A Unified Approach to Interpreting Model Predictions · Locator: Section 4.1, Kernel SHAP; Theorem 2 and endpoint constraints, PDF/printed pp. 5–6

⚠️ Curated Limitations & Negative Results

Results depend on how missing features are evaluated and on sampling; exact Shapley recovery in the full weighted problem is not a guarantee for every finite sample or background choice.
Independence approximations do not in general reproduce a dependence-aware conditional expectation.

Artificial Intelligence Ontology (AIO) Grounding

Mapped AIO Classes

AIO Target Class	Mapping Relation	Mapping Rationale
`AIO:Model` Model	related_match	The estimator analyzes a predictive model through feature-subset evaluations of its outputs. This direct analysis relationship does not identify the estimator with a model.

Evidence Sources & Primary Literature

1 Source

Source ID	Publication / Document	Type / Access	Reference & Checksum
`original-paper`	A Unified Approach to Interpreting Model Predictions Scott M. Lundberg, Su-In Lee	primary paper full_text	sha256:32518459c313…

Curation Provenance

Curator: Codex

Method: ai_assisted

Model: GPT-6

Created: 2026-10-11T04:02:04Z

Updated: 2026-10-11T04:02:04Z

Source File: data/curated/KernelSHAP.yaml