New methods accelerate Shapley value estimation for linear models

Researchers have developed three novel methods, two approximate and one exact, for efficiently estimating conditional Shapley values when employing a linear regression model as an explainer. These new techniques leverage constrained Gaussian Markov Random Field theory and sparse matrix algebra to jointly estimate coefficients of all submodels, significantly reducing computation time from hours to minutes or seconds. In numerical case studies, these methods demonstrated accuracy comparable to existing sequential approaches while offering substantial speed improvements, particularly when many coalitions need to be evaluated. AI

IMPACT Provides more efficient methods for model interpretability, potentially speeding up AI development and debugging.

RANK_REASON Academic paper detailing new statistical methodology. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• arXiv
• conditional Shapley values
• Fredrik Lohne Aanes
• linear explainer
• Shapr 3D

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →