mlsanityData to 14 Oct 2026

Explanations of Black-Box Models based on Directional Feature Interactions

arXiv 2304.07670 ↗First posted 16 Apr 2023cs.LGcited by 15

Not on the map. Why?

Abstract (arXiv)

As machine learning algorithms are deployed ubiquitously to a variety of domains, it is imperative to make these often black-box models transparent. Several recent works explain black-box models by capturing the most influential features for prediction per instance; such explanation methods are univariate, as they characterize importance per feature. We extend univariate explanation to a higher-order; this enhances explainability, as bivariate methods can capture feature interactions in black-box models, represented as a directed graph. Analyzing this graph enables us to discover groups of features that are equally important (i.e., interchangeable), while the notion of directionality allows us to identify the most influential features. We apply our bivariate method on Shapley value explanations, and experimentally demonstrate the ability of directional explanations to discover feature interactions. We show the superiority of our method against state-of-the-art on CIFAR10, IMDB, Census, Divorce, Drug, and gene data.

Key facts. Cited by 15 papers in the map (data through 14 Oct 2026). Builds on 31 of its 81 references that are in the map. First posted on arXiv 16 Apr 2023.

Builds on

31 of 81 references.

References · 31

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  4. 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,166
  5. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  6. 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,057
  7. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  8. 2019Fast Graph Representation Learning with PyTorch Geometriccited by 2,434
  9. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  10. 2019Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AIcited by 1,552
  11. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  12. 2017Towards A Rigorous Science of Interpretable Machine Learningcited by 1,416
19 more
  1. 2018A Survey Of Methods For Explaining Black Box Modelscited by 973
  2. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 838
  3. 2019GNNExplainer: Generating Explanations for Graph Neural Networkscited by 737
  4. 2020Parameterized Explainer for Graph Neural Networkcited by 426
  5. 2020Explainability in Graph Neural Networks: A Taxonomic Surveycited by 384
  6. 2018Consistent Individualized Feature Attribution for Tree Ensemblescited by 285
  7. 2021On Explainability of Graph Neural Networks via Subgraph Explorationscited by 281
  8. 2018Learning to Explain: An Information-Theoretic Perspective on Model Interpretationcited by 248
  9. 2019The Shapley Taylor Interaction Indexcited by 143
  10. 2019Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainabilitycited by 114
  11. 2020Explaining Explanations: Axiomatic Feature Interactions for Deep Networkscited by 103
  12. 2019CXPlain: Causal Explanations for Model Interpretation under Uncertaintycited by 98
  13. 2017Detecting Statistical Interactions from Neural Network Weightscited by 80
  14. 2020How does this interaction affect me? Interpretable attribution for feature interactionscited by 73
  15. 2021GraphSVX: Shapley Value Explanations for Graph Neural Networkscited by 64
  16. 2020Shapley Flow: A Graph-based Approach to Interpreting Model Predictionscited by 51
  17. 2020Feature Removal Is a Unifying Principle for Model Explanation Methodscited by 21
  18. 2018Can I trust you more? Model-Agnostic Hierarchical Explanationscited by 10
  19. 2020Scikit-network: Graph Analysis in Pythoncited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 6; 2024: 4; 2025: 4; 2026: 1.

15 papers.

Sort:

  1. 2025Exact Shapley Attributions in Quadratic-time for FANOVA Gaussian Processescited by 6
  2. 2023SmoothHess: ReLU Network Feature Interactions via Stein's Lemmacited by 5
  3. 2024Evaluating Input Feature Explanations through a Unified Diagnostic Evaluation Frameworkcited by 5
  4. 2025Amortized Linear-time Exact Shapley Value for Product-Kernel Methodscited by 5
  5. 2023Distributing Synergy Functions: Unifying Game-Theoretic Interaction Methods for Machine-Learning Explainabilitycited by 2
  6. 2024META-ANOVA: Screening interactions for interpretable machine learningcited by 2
  7. 2024Axiomatic Explainer Globalness via Optimal Transportcited by 2
  8. 2023Asymmetric feature interaction for interpreting model predictionscited by 1
  9. 2023Advancing Biomedicine with Graph Representation Learning: Recent Progress, Challenges, and Future Directionscited by 1
  10. 2023Explaining Interactions Between Text Spanscited by 1
  11. 2023DeLELSTM: Decomposition-based Linear Explainable LSTM to Capture Instantaneous and Long-term Effects in Time Seriescited by 0
  12. 2024PE: A Poincare Explanation Method for Fast Text Hierarchy Generationcited by 0
3 more
  1. 2025Exploring Theory-Laden Observations in the Brain Basis of Emotional Experiencecited by 0
  2. 2025OrdShap: Feature Position Importance for Sequential Black-Box Modelscited by 0
  3. 2026OMEGA: Optimizing Machine Learning by Evaluating Generated Algorithmscited by 0