mlsanityData to 7 Oct 2026

PredDiff: Explanations and Interactions from Conditional Expectations

arXiv 2102.13519 ↗First posted 26 Feb 2021cs.LG cs.AI stat.MLcited by 16

Not on the map. Why?

Abstract (arXiv)

PredDiff is a model-agnostic, local attribution method that is firmly rooted in probability theory. Its simple intuition is to measure prediction changes while marginalizing features. In this work, we clarify properties of PredDiff and its close connection to Shapley values. We stress important differences between classification and regression, which require a specific treatment within both formalisms. We extend PredDiff by introducing a new, well-founded measure for interaction effects between arbitrary feature subsets. The study of interaction effects represents an inevitable step towards a comprehensive understanding of black-box models and is particularly important for science applications. Equipped with our novel interaction measure, PredDiff is a promising model-agnostic approach for obtaining reliable, numerically inexpensive and theoretically sound attributions.

Key facts. Cited by 16 papers in the map (data through 7 Oct 2026). Builds on 36 of its 67 references that are in the map. First posted on arXiv 26 Feb 2021.

Builds on

36 of 67 references.

References · 36

  1. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  2. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  3. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  4. 2017On Calibration of Modern Neural Networkscited by 4,860
  5. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  6. 2013Visualizing and Understanding Convolutional Networkscited by 2,278
  7. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 440
  8. 2019Your Classifier is Secretly an Energy Based Model and You Should Treat it Like Onecited by 402
  9. 2019Unmasking Clever Hans Predictors and Assessing What Machines Really Learncited by 359
  10. 2017Methods for Interpreting and Understanding Deep Neural Networkscited by 342
  11. 2019The many Shapley values for model explanationcited by 341
  12. 2018Learning to Explain: An Information-Theoretic Perspective on Model Interpretationcited by 248
24 more
  1. 2020Explaining by Removing: A Unified Framework for Model Explanationcited by 218
  2. 2016Visualizing the Effects of Predictor Variables in Black Box Supervised Learning Modelscited by 212
  3. 2019Explaining individual predictions when features are dependent: More accurate approximations to Shapley valuescited by 205
  4. 2020Problems with Shapley-value-based explanations as feature importance measurescited by 198
  5. 2017Visualizing Deep Neural Network Decisions: Prediction Difference Analysiscited by 166
  6. 2019Feature relevance quantification in explainable AI: A causal problemcited by 164
  7. 2019The Shapley Taylor Interaction Indexcited by 143
  8. 2020Higher-Order Explanations of Graph Neural Networks via Relevant Walkscited by 138
  9. 2020Explaining Explanations: Axiomatic Feature Interactions for Deep Networkscited by 103
  10. 2020Benchmarking Deep Learning Interpretability in Time Series Predictionscited by 102
  11. 2020How does this interaction affect me? Interpretable attribution for feature interactionscited by 73
  12. 2020Fairwashing Explanations with Off-Manifold Detergentcited by 71
  13. 2018Visualizing the Feature Importance for Black Box Modelscited by 55
  14. 2020Interpretation of NLP models through input marginalizationcited by 43
  15. 2019Purifying Interaction Effects with the Functional ANOVA: An Efficient Algorithm for Recovering Identifiable Additive Modelscited by 41
  16. 2020Building and Interpreting Deep Similarity Modelscited by 34
  17. 2020Feature Removal Is a Unifying Principle for Model Explanation Methodscited by 21
  18. 2020Domain aware medical image classifier interpretation by counterfactual impact analysiscited by 17
  19. 2019Do Not Trust Additive Explanationscited by 15
  20. 2020A Unified Taylor Framework for Revisiting Attribution Methodscited by 13
  21. 2020Considering Likelihood in NLP Classification Explanations with Occlusion and Language Modelingcited by 11
  22. 2017Interpreting Classifiers through Attribute Interactions in Datasetscited by 8
  23. 2019Reluctant Interaction Modelingcited by 8
  24. 2019Contextual Prediction Difference Analysis for Explaining Individual Image Classificationscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 4; 2024: 7; 2025: 1; 2026: 2.

16 papers.

Sort:

  1. 2022From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagationcited by 141
  2. 2023Multi-dimensional concept discovery (MCD): A unifying framework with completeness guaranteescited by 48
  3. 2024Decoupling Pixel Flipping and Occlusion Strategy for Consistent XAI Benchmarkscited by 28
  4. 2024Unifying Feature-Based Explanations with Functional ANOVA and Cooperative Game Theorycited by 10
  5. 2024Towards Symbolic XAI -- Explanation Through Human Understandable Logical Relationships Between Featurescited by 7
  6. 2024Identifying Important Group of Pixels using Interactionscited by 5
  7. 2022Sparse Subspace Clustering for Concept Discovery (SSCCD)cited by 4
  8. 2023An XAI framework for robust and transparent data-driven wind turbine power curve modelscited by 4
  9. 2024Explaining Time Series Classification Predictions via Causal Attributionscited by 4
  10. 2024Explainable AI needs formalizationcited by 4
  11. 2024Explaining Object Detectors via Collective Contribution of Pixelscited by 3
  12. 2023Distributing Synergy Functions: Unifying Game-Theoretic Interaction Methods for Machine-Learning Explainabilitycited by 2
4 more
  1. 2023Insights Into the Inner Workings of Transformer Models for Protein Function Predictioncited by 2
  2. 2025Minimizing False-Positive Attributions in Explanations of Non-Linear Modelscited by 1
  3. 2026H-Sets: Hessian-Guided Discovery of Set-Level Feature Interactions in Image Classifierscited by 1
  4. 2026Explaining Temporal Graph Neural Networks via Feature-induced Information Flowcited by 0