Discovering additive structure in black box functions

By Giles Hooker. 2004

163 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Control and Systems Engineering, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

In Statistics and Probability

Citations in the corpus

Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “Others: [56, 219]”
A Survey of Methods for Explaining Black Box Models (2018, background): “In Reference [42], the authors present an approach f aimed at evaluating the importance of non-additive interactions between any set of features.”
GNNExplainer: Generating Explanations for Graph Neural Networks (2019, background): “Instead of creating new, inherently interpretable models, post-hoc interpretability methods [1, 14, 15, 17, 23, 38] consider models as black boxes and then probe them for relevant information.”
Estimating Mean Dimensionality of Analysis of Variance Decompositions (2006, background): “Then Υ2 u represents the cost, in lost model fidelity, of ignoring fu and fv for all v ⊇ u (Hooker 2004).”
Learning Global Additive Explanations for Neural Nets Using Model Distillation (2018, background): “, 2014), and decompose prediction functions into lower-dimensional components (Hooker, 2004).”
Survey of modeling and optimization strategies to solve high-dimensional design problems with computationally-expensive black-box functions (2010, background): “For example, Hooker (2004) discovered an additive structure; Chen (1991, 1993 ) made use of interactions; Owen (2000, 1998) discussed linearity in high dimensions.”
Explainable Deep Learning: A Field Guide for the Uninitiated (2020, background): “Global additive explanations (Sobol, 2001; Hooker, 2004; Hoos & Leyton-Brown, 2014) have been leveraged to study complex models, including analyzing how model parameters influence model performance and decomposing black box models into lower-dimensional components.”
Accurate intelligible models with pairwise interactions (2013, methodology): “Here we conduct experiments on synthetic data generated by the following function [14, 22].”
"Why Should You Trust My Explanation?" Understanding Uncertainty in LIME Explanations (2019, background): “1), individual conditional expectation plots (Goldstein et al., 2013), accumulated local effects plots (Apley, 2016), Friedman’s Hstatistic (Friedman et al., 2008), and variable interaction networks (Hooker, 2004). In the study of computer vision with neural networks, feature importance map (Iivarin”
Does My Multimodal Model Learn Cross-modal Interactions? It’s Harder to Tell than You Might Think! (2020, methodology): “Hooker (2004) gives a general method for computing the projection of a function f onto a set of functions with a specified ANOVA structure (see also Sobol (2001); Liu and Owen (2006)): our algorithmic contributions are to extend the method to multimodal models with d > 1 dimensional outputs, and to…”
Tree Space Prototypes: Another Look at Making Tree Ensembles Interpretable (2016, background): “nostic (not exclusive to tree ensembles) explanations of predictions [10, 11] and feature selection in tree ensembles using, for example, variable importance measures [12] or partial dependence plots [4, 13, 14, 15, 16]. The latter aims at interpretability by targeting sparsity-in-features [17], whe”
Explainable Artificial Intelligence: a Systematic Review (2020, methodology): “The Functional ANOVA decomposition [183] quantifies the influence of non-additive interactions within any set of input variables and depict them with Variable Interaction Network (VIN) graphs where the nodes represent the variables and the edges the interactions.”