Accurate intelligible models with pairwise interactions

By Yin Lou, R. Caruana, Johannes Gehrke, Giles Hooker. 2013

703 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability

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 Safety, Robustness & Interpretability

Citations in the corpus

Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, methodology): “If a GAM has bivariate component functions (that is, if we choose an fj to depend on two variables, which permits an interaction between these two variables), a heatmap can be used to visualize the component function on the two dimensional plane and understand the pairwise interactions (Lou et al., ”
The mythos of model interpretability (2016, background): “Our investigation of”
A Survey on Explainable Artificial Intelligence (XAI): Toward Medical XAI (2019, methodology): “tity. Modiﬁcations JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 7 can be duly implemented. As a natural extension to the model, interaction terms between variables f ij(x i;x j) are used [103]; we can certainly extend this indeﬁnitely. ProtoAttend [104] uses probabilities as weights in ”
Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead (2018, background): “, 8], causality, structural (generative) constraints, additivity [9], or physical constraints that come from domain knowledge.”
Please Stop Explaining Black Box Models for High Stakes Decisions (2018, background): “…is constrained in model form so that it is either useful to someone, or obeys structural knowledge of the domain, such as monotonicity [e.g., Gupta et al., 2016], causality, structural (generative) constraints, additivity [Lou et al., 2013], or physical constraints that come from domain knowledge.”
A Survey of Methods for Explaining Black Box Models (2018, methodology): “A refinement of the GAM approach is proposed by the same authors in Reference [70].”
Partially Interpretable Estimators (PIE): Black-Box-Refined Interpretable Machine Learning (2021, background): “We will compare with a classic GAM, an improved version of GA2M, called Explainable Boosted Machine (EBM) and the most recent NAM in this paper.”
Ethical Implementation of Artificial Intelligence to Select Embryos in In Vitro Fertilization (2021, background): “For tabular data, neural networks and other forms of black-box models do not seem to provide additional accuracy, which means that optimized medical scoring systems might be as accurate as one could get (depending on the dataset) [66, 67].”
Neural Additive Models: Interpretable Machine Learning with Neural Nets (2020, methodology): “Prior to NAMs, the state-of-the-art in high-accuracy, interpretable generalized additive models [12, 14] are the GAM [23] and GA2M [24] based on regularized boosted decision trees which were successfully applied to healthcare datasets [5].”
InterpretML: A Unified Framework for Machine Learning Interpretability (2019, methodology): “EBM is a fast implementation of the GA 2 M algorithm (Lou et al., 2013), written in C++ and Python.”
An Interpretable Model with Globally Consistent Explanations for Credit Risk (2018, methodology): “We did not require quadratic terms, as [8] did.”
Manipulating and Measuring Model Interpretability (2018, background): “simple models, such as point systems that can be memorized [35, 69] or generative additive models in which the impact of each feature on the model’s prediction should be relatively easy to understand [12, 50, 51]. The second is to provide post-hoc explanations for (potentially complex) models. One t”
Interpretable Decision Sets: A Joint Framework for Description and Prediction (2016, background): “For example, prototype vector machines [4], regression models [36, 35], generalized additive models [16, 40, 26, 27], Bayesian case models [19] mind-the-gap models [20], and scoring systems [37] have all been proposed as interpretable models.”
Causal Interpretability for Machine Learning - Problems, Methods and Evaluation (2020, background): “SIGKDD Explorations Volume 22, Issue 1 22 eralized Additive Models plus Interactions (GA(2)Ms) in the form of g(y) = ∑ fi(xi) + ∑ fij(xi, xj) which takes into account the two-dimensional interactions that still can be interpretable as heat maps [63].”
Explaining machine learning classifiers through diverse counterfactual explanations (2019, background): “Note that there has also been an important line of work that focuses on developing intelligible models by assuming that simple models such as linear models or decision trees are interpretable [7, 19, 21, 22].”
The ethics of algorithms: Mapping the debate (2016, neutral): “Further work is required to design broadly applicable, low impact auditing mechanisms for algorithms (cf. Adler et al., 2016; Sandvig et al., 2014) that build upon current work in transparency and interpretability of machine learning (e.g. Kim et al., 2015; Lou et al., 2013).”