mlsanityData to 14 Oct 2026

Concept Whitening for Interpretable Image Recognition

arXiv 2002.01650 ↗First posted 5 Feb 2020cs.LG cs.AI cs.CVcited by 225

Not on the map. Why?

Abstract (arXiv)

What does a neural network encode about a concept as we traverse through the layers? Interpretability in machine learning is undoubtedly important, but the calculations of neural networks are very challenging to understand. Attempts to see inside their hidden layers can either be misleading, unusable, or rely on the latent space to possess properties that it may not have. In this work, rather than attempting to analyze a neural network posthoc, we introduce a mechanism, called concept whitening (CW), to alter a given layer of the network to allow us to better understand the computation leading up to that layer. When a concept whitening module is added to a CNN, the axes of the latent space are aligned with known concepts of interest. By experiment, we show that CW can provide us a much clearer understanding for how the network gradually learns concepts over layers. CW is an alternative to a batch normalization layer in that it normalizes, and also decorrelates (whitens) the latent space. CW can be used in any layer of the network without hurting predictive performance.

Key facts. Cited by 225 papers in the map (data through 14 Oct 2026). First posted on arXiv 5 Feb 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 31; 2022: 47; 2023: 55; 2024: 45; 2025: 31; 2026: 16.

225 papers.

Sort:

By area:

  1. 2021Interpretable Machine Learning: Fundamental Principles and 10 Grand Challengescited by 367
  2. 2022Post-hoc Concept Bottleneck Modelscited by 293
  3. 2023Label-Free Concept Bottleneck Modelscited by 285
  4. 2023Getting aligned on representational alignmentcited by 143
  5. 2022From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagationcited by 141
  6. 2021Promises and Pitfalls of Black-Box Concept Learning Modelscited by 131
  7. 2022Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networkscited by 129
  8. 2021Acquisition of Chess Knowledge in AlphaZerocited by 123
  9. 2023Probabilistic Concept Bottleneck Modelscited by 101
  10. 2023Concept-based Explainable Artificial Intelligence: A Surveycited by 98
  11. 2021Interpretable Image Classification with Differentiable Prototypes Assignmentcited by 97
  12. 2021A Comprehensive Taxonomy for Explainable Artificial Intelligence: A Systematic Survey of Surveys on Methods and Conceptscited by 91
38 more
  1. 2021Transparency of Deep Neural Networks for Medical Image Analysis: A Review of Interpretability Methodscited by 83
  2. 2022From Concept Drift to Model Degradation: An Overview on Performance-Aware Drift Detectorscited by 83
  3. 2022Concept Bottleneck Model with Additional Unsupervised Conceptscited by 79
  4. 2022GlanceNets: Interpretabile, Leak-proof Concept-based Modelscited by 73
  5. 2023Learning Concise and Descriptive Attributes for Visual Recognitioncited by 73
  6. 2021IAIA-BL: A Case-based Interpretable Deep Learning Model for Classification of Mass Lesions in Digital Mammographycited by 72
  7. 2021Editing a classifier by rewriting its prediction rulescited by 65
  8. 2023Not All Neuro-Symbolic Concepts Are Created Equal: Analysis and Mitigation of Reasoning Shortcutscited by 63
  9. 2022Concept Embedding Models: Beyond the Accuracy-Explainability Trade-Offcited by 61
  10. 2022Post hoc Explanations may be Ineffective for Detecting Unknown Spurious Correlationcited by 59
  11. 2022Concept Activation Regions: A Generalized Framework For Concept-Based Explanationscited by 57
  12. 2023Interpretable Neural-Symbolic Concept Reasoningcited by 57
  13. 2022Leveraging Explanations in Interactive Machine Learning: An Overviewcited by 53
  14. 2022Concept-level Debugging of Part-Prototype Networkscited by 52
  15. 2021Entropy-based Logic Explanations of Neural Networkscited by 51
  16. 2023Multi-dimensional concept discovery (MCD): A unifying framework with completeness guaranteescited by 50
  17. 2022Beyond Explaining: Opportunities and Challenges of XAI-Based Model Improvementcited by 48
  18. 2021A Framework for Learning Ante-hoc Explainable Models via Conceptscited by 45
  19. 2021This looks more like that: Enhancing Self-Explaining Models by Prototypical Relevance Propagationcited by 43
  20. 2023Learning to Receive Help: Intervention-Aware Concept Embedding Modelscited by 41
  21. 2023Human Uncertainty in Concept-Based AI Systemscited by 39
  22. 2022Analyzing Encoded Concepts in Transformer Language Modelscited by 38
  23. 2023This Looks Like Those: Illuminating Prototypical Concepts Using Multiple Visualizationscited by 36
  24. 2021GCExplainer: Human-in-the-Loop Concept-based Explanations for Graph Neural Networkscited by 35
  25. 2022Guidelines and Evaluation of Clinical Explainable AI in Medical Image Analysiscited by 35
  26. 2023Auxiliary Losses for Learning Generalizable Concept-based Modelscited by 35
  27. 2023ProtoSeg: Interpretable Semantic Segmentation with Prototypical Partscited by 33
  28. 2023Neuro-Symbolic Continual Learning: Knowledge, Reasoning Shortcuts and Concept Rehearsalcited by 33
  29. 2022Explainable Deep Learning Methods in Medical Image Classification: A Surveycited by 32
  30. 2023Towards Robust Metrics for Concept Representation Evaluationcited by 29
  31. 2024Causal Concept Graph Models: Beyond Causal Opacity in Deep Learningcited by 29
  32. 2022Interpretable by Design: Learning Predictors by Composing Interpretable Queriescited by 28
  33. 2024Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?cited by 28
  34. 2023Interpretability Benchmark for Evaluating Spatial Misalignment of Prototypical Parts Explanationscited by 27
  35. 2023Coherent Concept-based Explanations in Medical Image and Its Application to Skin Lesion Diagnosiscited by 26
  36. 2022Navigating Neural Space: Revisiting Concept Activation Vectors to Overcome Directional Divergencecited by 25
  37. 2022The Free Energy Principle for Perception and Action: A Deep Learning Perspectivecited by 24
  38. 2023ICICLE: Interpretable Class Incremental Continual Learningcited by 24

50 most cited of 225 All 225 (JSON)