Study shows safely discarding features based on aggregate SHAP values is sound.
problem Discarding features based on aggregate SHAP values without proper justification.
method Investigated the soundness of discarding features based on aggregate SHAP values, proposing to aggregate SHAP values over the extended support.
result A small aggregate SHAP value implies safely discarding the corresponding feature.
X-SHAP assesses multiplicative variable contributions in machine learning models.
problem Understanding multiplicative interactions in machine learning models.
method Model-agnostic method that extends SHAP to assess multiplicative contributions.
result X-SHAP proves useful in capturing multiplicative feature importance.
This paper introduces glocal explanations for expected goal models in soccer.
problem Limited interpretability of expected goal models trained with black-box methods.
method Proposes glocal explanations using aggregated SHAP values and partial dependence profiles.
result Extracts knowledge from expected goal models for teams and players, enhancing performance analysis.
No feature ranking can be faithful, stable, and complete when features are collinear.
problem The impossibility of creating a feature ranking that is simultaneously faithful, stable, and complete when features are collinear.
method Proving the impossibility, quantifying it for four model classes, resolving it via ensemble averaging (DASH), and machine-verifying it with Lean 4 theorems.
result No method lies outside the dichotomy of faithful-complete methods (unstable, with rankings that flip up to 50% of the time) and ensemble methods (stable, reporting ties for symmetric features).
In this article we will analyse how to compute the contribution of each input value to its aggregate output in some nonlinear models. Regression and classification applications, together with related algorithms for deep neural networks are presented. The proposed approach merges two methods currently present in the lit…
Proposes simplified SHAP for faster black-box model explanations.
problem Computational expense of SHAP for models with many features.
method Ensemble of random SHAPs with feature selection and point generation.
result Efficiency and properties demonstrated through numerical experiments.
Paper proposes efficient SHAP computation methods.
problem Efficient computation of SHAP values for machine learning models.
method Develops polynomial time methods for SHAP computation based on model structure.
result Exact SHAP computation in polynomial time for various model structures.
SHAP explains boosted trees with additively modeled features.
problem Explaining predictions of boosted trees models with additively modeled features.
method SHAP values for additively modeled features in boosted trees models.
result SHAP dependence plot matches partial dependence plot for additively modeled features.
Proposes φ-table for statistical SHAP explanations in regression models.
problem Lack of clear directional summaries, uncertainty, and fidelity in SHAP feature importance.
method SHAP importance selection, fitting a standardized linear surrogate, reporting coefficients, uncertainty, fidelity, and stability.
result Extends SHAP into a statistical global explanation with direction, uncertainty, fidelity, and stability.
An imprecise SHAP method explains class probabilities with limited data.
problem Explaining class probabilities with limited training data.
method New approach for computing feature marginal contributions and general approach to interval-valued Shapley values.
result The imprecise SHAP method improves explanation of class probabilities.
Explaining complex or seemingly simple machine learning models is an important practical problem. We want to explain individual predictions from a complex machine learning model by learning simple, interpretable explanations. Shapley values is a game theoretic concept that can be used for this purpose. The Shapley valu…
The paper develops statistical inference methods for SHAP values.
problem Lack of statistical inference for SHAP values in model-agnostic feature importance.
method Semi-parametric approach using U-statistics and Neyman orthogonal scores for functionals of nested regressions.
result Asymptotically normal estimates of the pth powers of SHAP values for various p.
A neural network approach for efficient conditional SHAP calculations.
problem Efficiently calculating conditional SHAP values for various models.
method Surrogate neural network approach for conditional SHAP.
result Efficiently calculates conditional SHAP values for neural networks and other regression models.
This paper analyzes SHAP values using Fourier expansions for model interpretability.
problem Understanding and interpreting SHAP values in complex models.
method Developed a spectral framework using Fourier expansions for SHAP values in various model regimes.
result SHAP values are Lipschitz continuous in the deterministic regime and converge to Gaussian process values in the probabilistic regime.
The paper explores SHAP scores and their connection to functional ANOVA, highlighting challenges in approximations.
problem Estimating SHAP scores and understanding their limitations.
method Using the connection to functional ANOVA, the paper outlines the challenges in SHAP approximations and their relation to feature distribution and ANOVA terms.
result Challenges in SHAP approximations are primarily due to feature distribution and the number of ANOVA terms estimated.
G-SHAP generates multiple types of explanations for machine learning models.
problem Understanding model predictions and their differences across groups.
method Generalization of SHAP method to produce additional types of explanations.
result G-SHAP produces explanations for classification, intergroup differences, and model failure.
SHAP Distance assesses semantic fidelity of synthetic tabular data.
problem Semantic fidelity of synthetic tabular data is not well evaluated.
method SHAP Distance, defined as cosine distance between global SHAP attribution vectors.
result SHAP Distance detects semantic discrepancies overlooked by standard measures.
mSHAP explains predictions of two-part models, improving fairness and interpretability.
problem Interpreting predictions from two-part models, especially in insurance.
method mSHAP: a method for computing SHAP values of two-part models using the SHAP values of individual models.
result mSHAP is exponentially faster than kernelSHAP for computing approximate SHAP values.
CatNet controls FDR in LSTM models using SHAP feature importance and Gaussian mirrors.
problem Controlling False Discovery Rate (FDR) in LSTM models with feature selection.
method CatNet uses SHAP values for feature importance and Gaussian Mirror algorithm for FDR control. It introduces a kernel-based independence measure to handle feature correlations.
result CatNet reduces overfitting and improves model interpretability on simulated and real-world data.
New method explains survival analysis models using median-SHAP.
problem Need for explainable AI in medical applications, especially for survival analysis.
method Introduces median-SHAP for explaining survival analysis models.
result Conventionally used mean anchor point can lead to misleading interpretations; median-SHAP provides a better approach.
Study evaluates SHAP for credit card default model consistency.
problem Model transparency and fairness in credit card default prediction models.
method Evaluates SHAP stability in credit card default prediction models via a case study.
result SHAP consistency is related to variable importance level.
Neural networks predict US recessions with SHAP method.
problem Forecasting US recessions using machine learning.
method Long short-term memory (LSTM) and gated recurrent unit (GRU) models compared to linear models. SHAP method applied for interpretation.
result Neural networks can capture business cycle asymmetries and nonlinearities.
The paper uses SHAP for interpreting machine learning models in hospital data.
problem Interpreting machine learning models in healthcare.
method SHAP for feature importance and feature packing techniques.
result SHAP provides better interpretability of machine learning models in healthcare.
New method uses SHAP for biomarker identification in CATE models.
problem Identifying predictive biomarkers from observational data.
method Surrogate estimation approach using SHAP values for CATE meta-learners.
result SHAP accurately identifies biomarkers in high-dimensional data.
Instantly interprets black-box models using additive models.
problem Efficiency and representation power of SHAP explanations.
method Variational perspective linking GAM models and SHAP explanations; InstaSHAP method.
result Automatic computation of Shapley values in a single forward pass.
New framework for fairness in machine learning models using SHAP values and adversarial learning.
problem Fairness of model predictions, especially for unprivileged groups.
method Develops a new fairness definition and a framework using SHAP values and adversarial learning to mitigate bias.
result Models produced are fairer and performant, demonstrating the approach on various datasets.
SHAP clustering explains model predictions by grouping similar feature contributions.
problem Lack of explainability in black-box models.
method SHAP values for feature contributions, supervised clustering of SHAP values.
result Insight into pathways leading to similar predictions.
RoSHAP stabilizes feature attribution in machine learning models.
problem Stochastic variation in feature attribution measures.
method Modeling feature attribution score distribution and estimating it through bootstrap resampling and kernel density estimation.
result RoSHAP provides stable feature rankings and improves model performance.
Two XAI methods, SHAP and LIME, are discussed for tabular data models.
problem Making machine learning models transparent and trustworthy.
method SHAP and LIME methods for explaining model predictions.
result SHAP and LIME are model-dependent and sensitive to feature collinearity.
In the past years, many new explanation methods have been proposed to achieve interpretability of machine learning predictions. However, the utility of these methods in practical applications has not been researched extensively. In this paper we present the results of a human-grounded evaluation of SHAP, an explanation…
We solve the ANOVA decomposition for categorical inputs.
problem Lack of a closed-form expression for ANOVA decomposition with categorical dependent variables.
method Bridge functional analysis with discrete Fourier analysis to derive a closed-form decomposition.
result Closed-form decomposition for categorical inputs without assumptions.
Paper proposes a new method to evaluate AI model interpretability in bond default prediction.
problem Lack of standardized method to assess inherent interpretability of AI models.
method Uses LIME and SHAP to assess feature contributions in bond default prediction.
result Consistent results with intuitive understanding of model interpretability.
Unified framework for linear attribution methods in deep learning.
problem Separate theoretical foundations of XAI attribution methods.
method GRALIS (Gradient-Riesz Averaged Locally-Integrated Shapley) framework.
result Unified representation theory for linear attribution methods.
We discuss promising recent contributions on quantifying feature relevance using Shapley values, where we observed some confusion on which probability distribution is the right one for dropped features. We argue that the confusion is based on not carefully distinguishing between observational and interventional conditi…
Pref-SHAP explains preferences using Shapley values.
problem Challenging problem of preference explanation in machine learning.
method Shapley value-based model explanation framework for pairwise comparison data.
result Richer and more insightful explanations obtained over baseline.
Interpreting predictions from tree ensemble methods such as gradient boosting machines and random forests is important, yet feature attribution for trees is often heuristic and not individualized for each prediction. Here we show that popular feature attribution methods are inconsistent, meaning they can lower a featur…
Q-SHAP efficiently calculates feature contributions in boosting trees.
problem Global evaluation of feature contributions in tree models.
method Q-SHAP, an efficient algorithm that reduces Shapley values calculation to polynomial time.
result Q-SHAP improves computational efficiency and enhances accuracy of feature-specific R2 estimates. Two new algorithms speed up TreeSHAP computation for tree-based models.
problem Slow computation of SHAP values on tree-based models.
method Two new algorithms, Fast TreeSHAP v1 and v2, designed to improve computational efficiency.
result Fast TreeSHAP v2 is 2.5x faster than TreeSHAP, with slightly higher memory usage.
Study predicts soccer player market values using machine learning and SHAP for interpretability.
problem Predicting accurate market values for professional soccer players.
method Ensemble machine learning models, SHAP for interpretability, Boruta for feature selection.
result GBDT model achieved high predictive accuracy (R-squared 0.901, RMSE 3,221,632.175).
This research simplifies computation of feature attribution methods under certain conditions.
problem Computational complexity of feature attribution methods, especially power indices.
method Identifying conditions for polynomial computation and introducing new indices.
result Conditions for efficient computation of feature attribution methods are identified.
Machine learning predicts homicide clearance rates with SHAP explaining key features.
problem Predicting and explaining homicide clearance rates in the US.
method Nine algorithmic approaches compared; XGBoost selected. SHAP used for feature importance.
result XGBoost best predicts national homicide clearance rates; SHAP reveals key features.
XGBoost fails to accurately identify relevant features, while interpretable methods do.
problem Accurately identifying relevant features in black-box models like XGBoost.
method Comparison of variable importance methods (CART, Optimal Trees, XGBoost, SHAP) across various experiments.
result Interpretable methods outperform black-box models in feature selection accuracy.
A new tree-based estimator, FastPD, efficiently estimates PD functions for machine learning models.
problem Efficiently estimating Partial Dependence functions for machine learning models.
method Proposes a new tree-based estimator, FastPD, to estimate PD functions.
result FastPD consistently estimates the desired population quantity and improves complexity from quadratic to linear.
Unified local and global explanations through functional decomposition of low dimensional structures.
problem Tackles the challenge of extracting meaningful local and global explanations from machine learning models.
method Proposes a new identification constraint to decompose the global representation into main and interaction components of arbitrary order.
result Unified local and global explanations by connecting partial dependence plots and interventional SHAP values.
This paper enhances credit risk management using explainable AI techniques.
problem Lack of transparency and explainability in AI models for credit risk management.
method Implement LIME and SHAP for explaining ML-based credit scoring models.
result Demonstrates practical challenges and solutions for XAI methods in finance.
Paper describes anomaly detection and explainability for multivariate functional data.
problem Anomaly detection and explainability in multivariate functional data.
method Transform series into features, use Isolation Forest, compute SHAP coefficients, and use supervised decision tree.
result Method performs well on simulated and real industry data.
Conformal prediction fails under severe feature turnover in COVID-19 supply chain tasks.
problem Dealing with distribution shift in conformal prediction models.
method Using COVID-19 as a natural experiment across 8 supply chain tasks, analyzing SHAP explanations.
result Coverage drops vary widely (0% to 86.7%) and correlate with single-feature dependence.
As machine learning black boxes are increasingly being deployed in domains such as healthcare and criminal justice, there is growing emphasis on building tools and techniques for explaining these black boxes in an interpretable manner. Such explanations are being leveraged by domain experts to diagnose systematic error…