Hoeffding decomposition of black-box models with dependent inputs

arXiv:2310.06567v2 Announce Type: replace-cross
Abstract: One of the main challenges for interpreting black-box models is the ability to uniquely decompose square-integrable functions of non-independent random inputs into a sum of functions of every possible subset of variables. However, dealing with dependencies among inputs can be complicated. We propose a novel framework to study this problem, linking three domains of mathematics: probability theory, functional analysis, and combinatorics. We show that, under two reasonable assumptions on the inputs (non-perfect functional dependence and …

abstract arxiv box challenges dependencies every framework functions however independent inputs math.fa math.pr math.st novel random square stat.ml stat.th type variables