	US 12,423,614 B2
	Faithful and efficient sample-based model explanations
Yada Zhu, Irvington, NY (US); Wei Zhang, Acton, MA (US); Guangnan Ye, Scarsdale, NY (US); and Xiaodong Cui, Chappaqua, NY (US)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on May 31, 2021, as Appl. No. 17/334,889.
Prior Publication US 2022/0383185 A1, Dec. 1, 2022
Int. Cl. G06N 20/00 (2019.01); G06F 40/40 (2020.01)

CPC G06N 20/00 (2019.01) [G06F 40/40 (2020.01)]

19 Claims

1. A method for explaining a machine learning model θ, the method comprising:

training the machine learning model θ with training data D;

obtaining a decision of the machine learning model θ;

computing a set of faithful variants {θ_i} of the machine learning model θ using the training data D by:

randomly selecting batches B of the training data D;

calculating, for each batch B_ia gradient g (B_i|θ); and

computing the set of faithful variants {θ_i} of the machine learning model θ using the gradient g(B_i|θ) for each batch B_ias θ_i=θ+η_ig(B_i|θ), wherein η_iis an i-specific weighting parameter; and

explaining the decision of the machine learning model θ using a set of training examples from the set of faithful variants {θ_i} such that the decision is explained by a sum of influences of the set of faithful variants {θ_i}.