mlsanityData to 7 Oct 2026

Provable concept learning for interpretable predictions using variational autoencoders

arXiv 2204.00492 ↗First posted 1 Apr 2022cs.LG stat.MEcited by 9

Not on the map. Why?

Abstract (arXiv)

In safety-critical applications, practitioners are reluctant to trust neural networks when no interpretable explanations are available. Many attempts to provide such explanations revolve around pixel-based attributions or use previously known concepts. In this paper we aim to provide explanations by provably identifying high-level, previously unknown ground-truth concepts. To this end, we propose a probabilistic modeling framework to derive (C)oncept (L)earning and (P)rediction (CLAP) -- a VAE-based classifier that uses visually interpretable concepts as predictors for a simple classifier. Assuming a generative model for the ground-truth concepts, we prove that CLAP is able to identify them while attaining optimal classification accuracy. Our experiments on synthetic datasets verify that CLAP identifies distinct ground-truth concepts on synthetic datasets and yields promising results on the medical Chest X-Ray dataset.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 29 of its 37 references that are in the map. First posted on arXiv 1 Apr 2022.

Builds on

29 of 37 references.

References · 29

  1. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  2. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  3. 2017ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseasescited by 969
  4. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  5. 2018Disentangling by Factorisingcited by 614
  6. 2018Towards Robust Interpretability with Self-Explaining Neural Networkscited by 505
  7. 2014Semi-Supervised Learning with Deep Generative Modelscited by 494
  8. 2019Variational Autoencoders and Nonlinear ICA: A Unifying Frameworkcited by 494
  9. 2018Understanding disentangling in β-VAEcited by 332
  10. 2020StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generationcited by 331
  11. 2018On the Robustness of Interpretability Methodscited by 310
  12. 2021Self-Supervised Learning with Data Augmentations Provably Isolates Content from Stylecited by 308
17 more
  1. 2018Nonlinear ICA Using Auxiliary Variables and Generalized Contrastive Learningcited by 278
  2. 2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 240
  3. 2019On Completeness-aware Concept-Based Explanations in Deep Neural Networkscited by 239
  4. 2020Weakly-Supervised Disentanglement Without Compromisescited by 237
  5. 2020Problems with Shapley-value-based explanations as feature importance measurescited by 198
  6. 2018Explaining Image Classifiers by Counterfactual Generationcited by 144
  7. 2015An open access repository of images on plant health to enable the development of mobile disease diagnosticscited by 132
  8. 2021Explaining in Style: Training a GAN to explain a classifier in StyleSpacecited by 122
  9. 2019On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Datasetcited by 100
  10. 2020Getting a CLUE: A Method for Explaining Uncertainty Estimatescited by 92
  11. 2019The Incomplete Rosetta Stone Problem: Identifiability Results for Multi-View Nonlinear ICAcited by 86
  12. 2019Weakly Supervised Disentanglement with Guaranteescited by 82
  13. 2020Interpretability of machine learning based prediction models in healthcarecited by 75
  14. 2020Now You See Me (CME): Concept-based Model Extractioncited by 51
  15. 2018Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Insteadcited by 40
  16. 2020Capturing Label Characteristics in VAEscited by 30
  17. 2019Disentanglement Challenge: From Regularization to Reconstructioncited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 2; 2025: 3; 2026: 3.

9 papers.

  1. 2024A Neuro-Symbolic Benchmark Suite for Concept Quality and Reasoning Shortcutscited by 19
  2. 2025Shortcuts and Identifiability in Concept-based Models from a Neuro-Symbolic Lenscited by 13
  3. 2025Towards Interpretable Deep Generative Models via Causal Representation Learningcited by 12
  4. 2023Interpretable and intervenable ultrasonography-based machine learning models for pediatric appendicitiscited by 11
  5. 2025Nonparametric Identification of Latent Conceptscited by 3
  6. 2024Structural Causality-based Generalizable Concept Discovery Modelscited by 0
  7. 2026Concept Component Analysis: A Principled Approach for Concept Extraction in LLMscited by 0
  8. 2026Logit Distance Bounds Representational Similaritycited by 0
  9. 2026Beyond identifiability: Learning causal representations with few environments and finite samplescited by 0