English

Concept Whitening for Interpretable Image Recognition

Machine Learning 2020-12-14 v5 Artificial Intelligence Computer Vision and Pattern Recognition Machine Learning

Abstract

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.

Keywords

Cite

@article{arxiv.2002.01650,
  title  = {Concept Whitening for Interpretable Image Recognition},
  author = {Zhi Chen and Yijie Bei and Cynthia Rudin},
  journal= {arXiv preprint arXiv:2002.01650},
  year   = {2020}
}

Comments

Authors' pre-publication version of a 2020 Nature Machine Intelligence article

R2 v1 2026-06-23T13:31:36.119Z