mlsanityData to 7 Oct 2026

Concept-Based Techniques for "Musicologist-friendly" Explanations in a Deep Music Classifier

arXiv 2208.12485 ↗First posted 26 Aug 2022cs.SD cs.AI eess.AScited by 12

Not on the map. Why?

Abstract (arXiv)

Current approaches for explaining deep learning systems applied to musical data provide results in a low-level feature space, e.g., by highlighting potentially relevant time-frequency bins in a spectrogram or time-pitch bins in a piano roll. This can be difficult to understand, particularly for musicologists without technical knowledge. To address this issue, we focus on more human-friendly explanations based on high-level musical concepts. Our research targets trained systems (post-hoc explanations) and explores two approaches: a supervised one, where the user can define a musical concept and test if it is relevant to the system; and an unsupervised one, where musical excerpts containing relevant concepts are automatically selected and given to the user for interpretation. We demonstrate both techniques on an existing symbolic composer classification system, showcase their potential, and highlight their intrinsic limitations.

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Builds on 27 of its 39 references that are in the map. First posted on arXiv 26 Aug 2022.

Builds on

27 of 39 references.

References · 27

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  3. 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
  4. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
  5. 2017Explanation in Artificial Intelligence: Insights from the Social Sciencescited by 1,201
  6. 2018Sanity Checks for Saliency Mapscited by 1,149
  7. 2017Understanding the Effective Receptive Field in Deep Convolutional Neural Networkscited by 549
  8. 2017Interpretation of Neural Networks is Fragilecited by 443
  9. 2017The (Un)reliability of saliency methodscited by 331
  10. 2020Concept Whitening for Interpretable Image Recognitioncited by 225
  11. 2018Enabling Factorized Piano Music Modeling and Generation with the MAESTRO Datasetcited by 215
  12. 2014Object Detectors Emerge in Deep Scene CNNscited by 178
15 more
  1. 2018Activation Functions: Comparison of trends in Practice and Research for Deep Learningcited by 130
  2. 2020Invertible Concept-based Explanations for CNN Models with Non-negative Concept Activation Vectorscited by 119
  3. 2016TensorLy: Tensor Learning in Pythoncited by 104
  4. 2014Convolutional Feature Masking for Joint Object and Stuff Segmentationcited by 41
  5. 2022Listen to Interpret: Post-hoc Interpretability for Audio Networks with NMFcited by 23
  6. 2020audioLIME: Listenable Explanations Using Source Separationcited by 22
  7. 2021Prototype Learning for Interpretable Respiratory Sound Analysiscited by 21
  8. 2020Reliable Local Explanations for Machine Listeningcited by 17
  9. 2021Robust Semantic Interpretability: Revisiting Concept Activation Vectorscited by 15
  10. 2019Towards Explainable Music Emotion Recognition: The Route via Mid-level Featurescited by 10
  11. 2020Large-Scale MIDI-based Composer Classificationcited by 10
  12. 2020Deep Composer Classification Using Symbolic Representationcited by 9
  13. 2022Interpretability for Multimodal Emotion Recognition using Concept Activation Vectorscited by 8
  14. 2020Towards Musically Meaningful Explanations Using Source Separationcited by 4
  15. 2021On the Veracity of Local, Model-agnostic Explanations in Audio Classification: Targeted Investigations with Adversarial Examplescited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 5; 2025: 3; 2026: 2.

12 papers.

  1. 2024Explainable and Interpretable Multimodal Large Language Models: A Comprehensive Surveycited by 52
  2. 2024Detecting music deepfakes is easy but actually hardcited by 7
  3. 2023Tackling Interpretability in Audio Classification Networks with Non-negative Matrix Factorizationcited by 5
  4. 2024MusicLIME: Explainable Multimodal Music Understandingcited by 3
  5. 2023Predicting Music Hierarchies with a Graph-Based Neural Decodercited by 2
  6. 2024Prevailing Research Areas for Music AI in the Era of Foundation Modelscited by 1
  7. 2024PBSCR: The Piano Bootleg Score Composer Recognition Datasetcited by 0
  8. 2025Semantic-Aware Interpretable Multimodal Music Auto-Taggingcited by 0
  9. 2025Style-based Composer Identification and Attribution of Symbolic Music Scores: a Systematic Surveycited by 0
  10. 2025MUSE-Explainer: Counterfactual Explanations for Symbolic Music Graph Classification Modelscited by 0
  11. 2026ConceptCaps: a Distilled Concept Dataset for Interpretability in Music Modelscited by 0
  12. 2026Uncovering shortcut learning in audio classifiers by discovering recurring concepts in temporal explanationscited by 0