Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Leave-One-EquiVariant: Alleviating invariance-related information loss in contrastive music representations

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A contrastive music encoder can be trained to stay equivariant to pitch shift and time stretch by giving each augmentation its own projection head, so that attribute information survives in the frozen embeddings instead of being erased.

desk verdict A fair adaptation of LOOC to music with a useful tracking scheme; the empirical case is plausible, but the variant-head positive set has an acknowledged soft spot that needs tightening. read the letter →

arxiv 2412.18955 v1 pith:JJYPBPZY submitted 2024-12-25 cs.SD eess.AS

Julien Guinot , Elio Quinton , György Fazekas This is my paper · ORCID
classification cs.SDeess.AS
keywords contrastivelearningmusicrepresentationequivariancelearnedinvariancespitchshiftingtimestretchingdisentangledlatentspaceinformationretrieval
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that contrastive music representations do not have to pay the usual price for augmentation robustness. It proposes Leave-One-EquiVariant (LOEV): alongside the standard all-invariant projection head, add one head for each augmentation whose attribute matters downstream, and define that head's positives as views that did not receive that augmentation. The shared encoder must then preserve the augmentation's information, so key, pitch, and tempo survive in frozen embeddings while tagging stays unaffected. A second variant, LOEV++, structurally disentangles the latent space into invariant, pitch-variant, and tempo-variant subspaces, enabling retrieval by those attributes.

What carries the argument

The central mechanism is the variant positive set combined with a per-sample augmentation tracking vector. Each sample i carries a binary vector ti indicating which of the K variant augmentations were applied, and for augmentation Tk the positive set is Pk(i) = P(i) ∩ {j | tj,k = ti,k = 0}, so only views that both avoided Tk are pulled together. This forces the k-th projection head to stay equivariant to Tk while remaining invariant to everything else, and the global loss averages the all-invariant head with all variant heads. LOEV++ strengthens the scheme by parallelizing part of the last encoder block into each head, creating separate subspaces Vi, Vp, and Vt that can be probed or concatenated as V++.

What would settle it

Train LOEV with pitch-shift amounts drawn from a tiny interval (e.g., ±0.05 semitones) so two independently shifted views are nearly identical; if key-estimation probing then collapses toward the all-invariant baseline, the separation of the variant-head positive set rests on continuous-parameter diversity rather than on a robust principle.

Watch

Extended reading notes

Core claim

The paper's central claim is that a contrastive encoder can be made to retain attribute information that standard all-invariant training discards, by allocating one extra projection head per augmentation of interest and defining that head's positives to be views that did NOT receive that augmentation. The shared encoder then must keep the augmentation's information in order to satisfy all heads, producing an embedding superspace V from which pitch-key and tempo can still be probed and retrieved. The paper further claims that LOEV++, which resorbs part of the encoder into the heads, yields disentangled subspaces — an invariant space Vi, a pitch-variant space Vp, and a stretch-variant space Vt — with targeted retrieval working best in the corresponding variant space. The reported evidence: on key estimation, LOEV++ PSTS reaches 44.2 weighted accuracy versus 15.1 for MULE++; tempo acc1/acc2 reach 72.6/91.1 versus 63.2/84.1; automatic tagging stays at baseline (MTAT AUROC 90.6).

Load-bearing premise

The method's load-bearing premise is that two views that both received the target augmentation are never so similar that they should be positives, because continuous parameters are sampled uniformly; if the sampled shifts or stretch factors coincide or are perceptually identical, the variant head is pushed to pull genuine positives apart and the preserved-information objective degrades.

Editorial extensions

If this is right

  • Adding pitch shifting to pretraining no longer sacrifices key and pitch information: LOEV++ with pitch and stretch variants reaches 44.2 weighted key accuracy on Giantsteps versus 15.1 for MULE++ trained with the same augmentations.
  • Tempo information survives time stretching: LOEV++ PSTS reaches 72.6/91.1 acc1/acc2 on the AllTempo benchmark versus 63.2/84.1 for MULE++.
  • Automatic tagging is not harmed: MTG-Jamendo and MagnaTagATune AUROC and average precision remain at the MULE++ baseline level.
  • LOEV++ disentangles the latent space by design: key retrieval is best in the pitch-variant subspace Vp, tempo retrieval is best in the stretch-variant subspace Vt, enabling attribute-targeted retrieval.
  • The framework is transformation-agnostic, so any augmentation with a semantic downstream attribute can be given its own variant head without changing the training objective's structure.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension is to add same-parameter-augmented pairs as an extra positive set for the variant head, which would directly address the paper's assumption that continuous parameter sampling guarantees distinct views and could tighten the preserved-information objective.
  • The one-head-per-augmentation cost grows linearly with the number of variant augmentations, and sparse contrastive matrices become a practical limit; this suggests a selection criterion for which augmentations deserve variant heads, a question the paper leaves open.
  • The same positive-exclusion scheme could be applied to nonparametric semantic transformations such as genre or instrumentation if a continuous proxy parameter can be defined, potentially extending equivariance beyond audio effects.
  • Because the paper only evaluates 3-second chunks, the method's ability to preserve key and tempo at track level — where key changes and tempo drifts occur — remains untested and is a natural next benchmark.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes Leave-One-EquiVariant (LOEV) and its variant LOEV++ to reduce invariance-related information loss in contrastive music representations. In addition to the standard all-invariant projection head, LOEV trains one extra projection head per selected augmentation, with the positive set for head k restricted to pairs in which neither view received augmentation Tk. The authors argue that this forces the shared encoder to preserve information about all tracked augmentations. LOEV++ additionally resorbs part of the last encoder block into the parallel projection heads to create a concatenated representation space. The paper evaluates frozen representations from models pretrained on MTG-Jamendo on automatic tagging, key estimation, pitch estimation, tempo estimation, and retrieval tasks. The main claimed results are that LOEV and LOEV++ recover much of the key and tempo accuracy lost by standard MULE when pitch shifting and time stretching are used, without harming tagging, and that LOEV++ yields somewhat specialized subspaces for pitch and tempo information.

Significance. If supported, the method would be a practical and reasonably light-weight way to make contrastive music representations task-adaptive, addressing a real limitation of augmentation chains. The paper also transfers the LOOC idea from vision to music with an explicit link to semantically meaningful attributes (key and tempo), which is a useful contribution. The appendices are unusually candid about design limitations, and the paper ships a substantial set of experiments spanning probing and retrieval. However, the empirical support is weakened by the absence of error bars or repeated runs, by per-dataset tuning of probe architectures, and by an underspecified positive-set mechanism that the authors themselves partially question in Appendix D. These issues are fixable and do not, by themselves, invalidate the contribution, but they currently prevent the strong claims in the conclusion from being fully supported.

major comments (3)
  1. [II-B and Appendix D] The construction Pk(i) = P(i) ∩ {j | tj,k = ti,k = 0} means that when an anchor has received the target augmentation Tk, its positive set is empty, so the variant head receives no direct positive supervision for transformed views. As a result, the head may learn only a binary separation between Tk=0 and Tk=1 rather than the continuous shift or stretch parameter that key and tempo probing require. In addition, pairs that both received Tk with nearly equal sampled parameters are treated as negatives, relying on the assertion that uniform sampling of continuous parameters guarantees distinct embeddings. Appendix D explicitly concedes that perceptually similar augmented samples are treated the same as radically different augmented samples, which could introduce confusion into the training objective. The authors should provide an analysis or ablation demonstrating that the variant head encodes the continuous augmentation parameter and that near-identical transformed pairs do not degrade the preserved-information objective.
  2. [III-C, Table I and Conclusion] All reported results come from single pretraining and probing runs, with no error bars, and probe architectures, dropout, and learning rates are empirically adjusted per dataset. The conclusion's central claim that LOEV(++) 'reduces information loss ... with no detriment to tagging performance' rests on small tagging differences, e.g., MTAT AUROC 90.6 vs 90.5 and AP 38.4 vs 38.7 between LOEV-PSTS and MULE-PSTS, which cannot be distinguished from run-to-run noise. The authors should report means and standard deviations over multiple seeds, and ideally statistical significance, for the key comparisons: LOEV-PSTS vs MULE-PSTS and LOEV++ vs MULE++ on tagging, key, pitch, and tempo tasks.
  3. [III-D, Table II and Fig. 3] The claim that LOEV++ creates a disentangled latent space is supported only by point estimates and qualitative inspection. In Table II, the pitch subspace gives key accuracy 43.0 versus 39.0 for the invariant subspace and 30.0 for the tempo subspace, and the tempo subspace gives tempo acc1 71.5 versus 64.7 for the pitch subspace; Fig. 3 shows cosine-distance curves without uncertainty. Given that these differences motivate the disentanglement claim, the authors should add a quantitative disentanglement evaluation with confidence intervals, or at least repeated runs, rather than relying on single-probe point estimates.
minor comments (4)
  1. [Appendix D] The word 'defauly' should be 'default' in the in-track sampling strategy paragraph.
  2. [Table IV] In the time-stretching row, the parameter name 'transpose' should be 'stretch factor'; the current wording is confusing.
  3. [III-A] The probe time-stretching augmentation is specified as tau ~ U(0.8, 1.2), while the pretraining time-stretch range in Table IV is 0.7 to 1.3; the relationship between these two ranges should be clarified.
  4. [II-B] The notation Pk(i) = P(i) ∩ {j | tj,k = ti,k = 0} is clear in symbols, but the preceding sentence says only 'samples from the same anchor that have not been augmented with Tk'; the requirement that the anchor itself also have ti,k = 0 should be stated explicitly to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: LOEV(++)'s variant heads are trained with a self-supervised objective and validated on external held-out labels.

full rationale

The central claim is empirical and self-contained. LOEV's variant projection heads are trained with Eq. (2), a self-supervised contrastive loss whose positive set excludes pairs that both received the target augmentation; no downstream labels, key/tempo ground truth, or retrieval targets enter pretraining. The subsequent evaluation probes frozen embeddings on external datasets (Giantsteps, NSynth, AllTempo, MTAT) with held-out labels, so the reported gains in key, pitch, tempo, and retrieval are measured outcomes rather than quantities fitted by the method. The only load-bearing assumption, that stochastic uniform sampling of continuous augmentation parameters makes two same-Tk views non-identical, is explicitly flagged in Appendix D: 'perceptually-similar augmented samples are treated the same as radically different augmented samples with our strategy, which might introduce confusion into the training objective'; this is a robustness limitation of the positive-set design, not a circular reduction. The two self-citations ([8], [32]) are related-work and probe-detail citations and carry no load-bearing argument, and no uniqueness theorem or ansatz is imported from the authors' prior work. The method is an acknowledged adaptation of external LOOC [16] and is evaluated against similarly scaled baselines, so the derivation chain does not reduce to its own inputs.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim is empirical and rests on standard supervised and probing evaluation assumptions. No new physical entities are introduced. The main free parameters are design choices and per-dataset probe settings. The two domain assumptions listed are explicitly stated by the authors and are load-bearing for the variant-head training; if either fails, the heads receive corrupted positive or negative signal.

free parameters (3)
  • Variant augmentation application probability = 0.5 (pitch shift), 0.5 (time stretch)
    Set by hand; the appendix notes that higher probabilities create sparse or empty contrastive matrices, and no ablation over this value is reported.
  • Projection head size = two 2048-wide hidden layers
    Scaled down from MULE because multiple heads are used; no ablation supporting the size is reported.
  • Probe hyperparameters = per-dataset depth, width, dropout, learning rate
    Empirically adjusted for each dataset in Section III-C; this tuning is not tied to a fixed validation protocol in the preprint, so part of the reported differences may come from probe fitting.
assumptions (3)
  • domain assumption Key and tempo are approximately position-invariant within a music track, so sampling positive chunks from different positions does not change the semantic attributes the variant heads try to preserve.
    Invoked in Section II-C to justify chunk sampling strategies; if a track has local key changes, same-track positives can become false positives for the pitch-variant head.
  • ad hoc to paper Uniform random sampling of continuous augmentation parameters makes two views that both received the target augmentation sufficiently different to be treated as negatives.
    Section II-B; this is what permits excluding both-augmented pairs from positive sets. If two samples receive nearly identical parameter values, the model is asked to separate near-duplicates.
  • ad hoc to paper Resorbing a portion of the last encoder block into parallel heads produces subspaces that a linear probe can read as disentangled.
    Appendix C acknowledges disentanglement is modest; the architecture choice and its sufficiency are not derived, only demonstrated in Table II.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leave-One-EquiVariant: Alleviating invariance-related information loss in contrastive music representations." pith.science (2026). https://pith.science/paper/JJYPBPZY

@misc{pith2026241218955,
  author       = {Pith},
  title        = {Pith review of: Leave-One-EquiVariant: Alleviating invariance-related information loss in contrastive music representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JJYPBPZY}},
  note         = {Machine review of arXiv:2412.18955}
}
read the original abstract

Contrastive learning has proven effective in self-supervised musical representation learning, particularly for Music Information Retrieval (MIR) tasks. However, reliance on augmentation chains for contrastive view generation and the resulting learnt invariances pose challenges when different downstream tasks require sensitivity to certain musical attributes. To address this, we propose the Leave One EquiVariant (LOEV) framework, which introduces a flexible, task-adaptive approach compared to previous work by selectively preserving information about specific augmentations, allowing the model to maintain task-relevant equivariances. We demonstrate that LOEV alleviates information loss related to learned invariances, improving performance on augmentation related tasks and retrieval without sacrificing general representation quality. Furthermore, we introduce a variant of LOEV, LOEV++, which builds a disentangled latent space by design in a self-supervised manner, and enables targeted retrieval based on augmentation related attributes.

Figures

Figures reproduced from arXiv: 2412.18955 by the authors.

Figure 1
Figure 1. Leave One EquiVariant framework. Subspace [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Retrieval metrics for retrieved tags (MagnaTagATune), key (Giantsteps), and tempo (AllTempo) - Precision@K, Weighted accuracy [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Cosine distance between embeddings of pitch-shifted and non [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: LOEV(++) architectures. In either case, the probing represen [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Balancing Information Preservation and Disentanglement in Self-Supervised Music Representation Learning

    cs.SD 2025-07 conditional novelty 5.0 of 10

    A multi-view SSL framework with combined reconstruction and separation-based contrastive losses obtains disentangled pitch and instrument subspaces without the accuracy loss seen with contrastive-only training on NSynth.

Reference graph

Works this paper leans on

33 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [1]

    A simple framework for contrastive learning of visual representations,

    Ting Chen, Simon Kornblith, Mohammad Norouzi, et al., “A simple framework for contrastive learning of visual representations,” in Inter- national conference on machine learning. PMLR, 2020, pp. 1597–1607

    work page 2020

  2. [2]

    Contrastive Learning of Musical Representations,

    Janne Spijkervet and John Ashley Burgoyne, “Contrastive Learning of Musical Representations,” in International Society for Music Information Retrieval (ISMIR). Sept. 2021, number arXiv:2103.09410, arXiv

    arXiv 2021

  3. [3]

    S3t: Self-supervised pre- training with swin transformer for music classification,

    Hang Zhao, Chen Zhang, Bilei Zhu, et al., “S3t: Self-supervised pre- training with swin transformer for music classification,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2022, pp. 606–610

    work page 2022

  4. [4]

    Tailed u-net: Multi-scale music representation learning.,

    Marcel A V ´elez V ´asquez and John Ashley Burgoyne, “Tailed u-net: Multi-scale music representation learning.,” in ISMIR, 2022, pp. 67–75

    work page 2022

  5. [5]

    Supervised and Unsupervised Learning of Audio Representations for Music Understanding,

    Matthew C. McCallum, Filip Korzeniowski, Sergio Oramas, et al., “Supervised and Unsupervised Learning of Audio Representations for Music Understanding,” in International Society for Music Information Retrieval (ISMIR). Oct. 2022, number arXiv:2210.03799, arXiv

    arXiv 2022

  6. [6]

    Multi-Source Contrastive Learning from Musical Audio

    Christos Garoufis, Athanasia Zlatintsi, and Petros Maragos, “Multi- Source Contrastive Learning from Musical Audio,” May 2023, number arXiv:2302.07077, arXiv

    work page Pith review arXiv 2023

  7. [7]

    To- wards proper contrastive self-supervised learning strategies for music audio representation,

    Jeong Choi, Seongwon Jang, Hyunsouk Cho, and Sehee Chung, “To- wards proper contrastive self-supervised learning strategies for music audio representation,” in 2022 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2022, pp. 1–6

    work page 2022

  8. [8]

    Semi-Supervised Contrastive Learning of Musical Representations

    Julien Guinot, Elio Quinton, and Gy ¨orgy Fazekas, “Semi-supervised contrastive learning of musical representations,” arXiv preprint arXiv:2407.13840, 2024

    work page Pith review arXiv 2024

Show all 33 references
  1. [9]

    Contrastive Learning with Positive-Negative Frame Mask for Music Representation,

    Dong Yao, Zhou Zhao, Shengyu Zhang, et al., “Contrastive Learning with Positive-Negative Frame Mask for Music Representation,” in Proceedings of the ACM Web Conference 2022 , Apr. 2022, pp. 2906– 2915

    2022

  2. [10]

    COCOLA: Coherence-Oriented Contrastive Learning of Musical Audio Representations,

    Ruben Ciranni, Emilian Postolache, Giorgio Mariani, Michele Mancusi, Luca Cosmo, and Emanuele Rodol `a, “COCOLA: Coherence-Oriented Contrastive Learning of Musical Audio Representations,” Apr. 2024

    2024

  3. [11]

    Class-Aware Contrastive Semi-Supervised Learning,

    Fan Yang, Kai Wu, Shuyi Zhang, Guannan Jiang, Yong Liu, Feng Zheng, Wei Zhang, Chengjie Wang, and Long Zeng, “Class-Aware Contrastive Semi-Supervised Learning,” Sept. 2022

    2022

  4. [12]

    Boosting Contrastive Self-Supervised Learning with False Negative Cancellation,

    Tri Huynh, Simon Kornblith, Matthew R. Walter, Michael Maire, and Maryam Khademi, “Boosting Contrastive Self-Supervised Learning with False Negative Cancellation,” in 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , Waikoloa, HI, USA, Jan. 2022, pp. ...

    2022

  5. [13]

    Robust Contrastive Learning Using Negative Samples with Diminished Semantics,

    Songwei Ge, Shlok Mishra, Chun-Liang Li, Haohan Wang, and David Jacobs, “Robust Contrastive Learning Using Negative Samples with Diminished Semantics,” in Advances in Neural Information Processing Systems. 2021, vol. 34, pp. 27356–27368, Curran Associates, Inc

    2021

  6. [14]

    Wav2CLIP: Learning Robust Audio Representations from Clip,

    Ho-Hsiang Wu, Prem Seetharaman, Kundan Kumar, et al., “Wav2CLIP: Learning Robust Audio Representations from Clip,” in ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), May 2022, pp. 4563–4567

    2022

  7. [15]

    What Makes for Good Views for Contrastive Learning?,

    Yonglong Tian, Chen Sun, Ben Poole, Dilip Krishnan, Cordelia Schmid, and Phillip Isola, “What Makes for Good Views for Contrastive Learning?,” in Advances in Neural Information Processing Systems . 2020, vol. 33, pp. 6827–6839, Curran Associates, Inc

    2020

  8. [16]

    What Should Not Be Contrastive in Contrastive Learning,

    Tete Xiao, Xiaolong Wang, Alexei A. Efros, and Trevor Darrell, “What Should Not Be Contrastive in Contrastive Learning,” Mar. 2021

    2021

  9. [17]

    On the effect of data-augmentation on local embedding properties in the contrastive learning of music audio representations,

    Matthew C McCallum, Matthew EP Davies, Florian Henkel, Jaehun Kim, and Samuel E Sandberg, “On the effect of data-augmentation on local embedding properties in the contrastive learning of music audio representations,” in ICASSP 2024-2024 IEEE International Conference on Acousti...

    2024

  10. [18]

    Improved Baselines with Momentum Contrastive Learning,

    Xinlei Chen, Haoqi Fan, Ross Girshick, et al., “Improved Baselines with Momentum Contrastive Learning,” Mar. 2020

    2020

  11. [19]

    Towards Learning Universal Audio Representations,

    Luyu Wang, Pauline Luc, Yan Wu, et al., “Towards Learning Universal Audio Representations,” in ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , May 2022, pp. 4593–4597

    2022

  12. [20]

    The mtg- jamendo dataset for automatic music tagging,

    Dmitry Bogdanov, Minz Won, Philip Tovstogan, et al., “The mtg- jamendo dataset for automatic music tagging,” in Machine Learning for Music Discovery Workshop, International Conference on Machine Learning (ICML 2019) , Long Beach, CA, United States, 2019

    2019

  13. [21]

    Evaluation of algorithms using games: The case of music tagging.,

    Edith Law, Kris West, Michael I Mandel, et al., “Evaluation of algorithms using games: The case of music tagging.,” in ISMIR. Citeseer, 2009, pp. 387–392

    2009

  14. [22]

    End-to-end learning for music audio tagging at scale,

    Jordi Pons, Oriol Nieto, Matthew Prockup, Erik Schmidt, Andreas Ehmann, and Xavier Serra, “End-to-end learning for music audio tagging at scale,” arXiv preprint arXiv:1711.02520 , 2017

    2017 arXiv

  15. [23]

    Two data sets for tempo estimation and key detection in electronic dance music annotated from user corrections,

    Peter Knees, ´Angel Faraldo P ´erez, Herrera Boyer, Richard V ogl, Se- bastian B ¨ock, Florian H ¨orschl¨ager, Mickael Le Goff, et al., “Two data sets for tempo estimation and key detection in electronic dance music annotated from user corrections,” in Proceedings of the 16th ...

    2015

  16. [24]

    End-to-end musical key estimation using a convolutional neural network,

    Filip Korzeniowski and Gerhard Widmer, “End-to-end musical key estimation using a convolutional neural network,” in2017 25th European Signal Processing Conference (EUSIPCO) . IEEE, 2017, pp. 966–970

    2017

  17. [25]

    MARBLE: Music Audio Representation Benchmark for Universal Evaluation,

    Ruibin Yuan, Yinghao Ma, Yizhi Li, et al., “MARBLE: Music Audio Representation Benchmark for Universal Evaluation,” July 2023

    2023

  18. [26]

    Mir eval: A transparent implementation of common mir metrics.,

    Colin Raffel, Brian McFee, Eric J Humphrey, Justin Salamon, Oriol Nieto, Dawen Liang, Daniel PW Ellis, and C Colin Raffel, “Mir eval: A transparent implementation of common mir metrics.,” in ISMIR, 2014, vol. 10, p. 2014

    2014

  19. [27]

    Neural audio synthesis of musical notes with wavenet autoencoders,

    Jesse Engel, Cinjon Resnick, Adam Roberts, Sander Dieleman, Mo- hammad Norouzi, Douglas Eck, and Karen Simonyan, “Neural audio synthesis of musical notes with wavenet autoencoders,” in International Conference on Machine Learning . PMLR, 2017, pp. 1068–1077

    2017

  20. [28]

    Musical genre classification of audio signals,

    George Tzanetakis and Perry Cook, “Musical genre classification of audio signals,” IEEE Transactions on speech and audio processing , vol. 10, no. 5, pp. 293–302, 2002

    2002

  21. [29]

    Perceptual tempo estima- tion using gmm-regression,

    Geoffroy Peeters and Joachim Flocon-Cholet, “Perceptual tempo estima- tion using gmm-regression,” in Proceedings of the second international ACM workshop on Music information retrieval with user-centered and multimodal strategies, 2012, pp. 45–50

    2012

  22. [30]

    Particle filtering applied to musical tempo tracking,

    Stephen W Hainsworth and Malcolm D Macleod, “Particle filtering applied to musical tempo tracking,” EURASIP Journal on Advances in Signal Processing, vol. 2004, pp. 1–11, 2004

    2004

  23. [31]

    Similar but faster: manipulation of tempo in music audio embeddings for tempo prediction and search,

    Matthew C McCallum, Florian Henkel, Jaehun Kim, Samuel E Sandberg, and Matthew EP Davies, “Similar but faster: manipulation of tempo in music audio embeddings for tempo prediction and search,” arXiv preprint arXiv:2401.08902, 2024

    2024 arXiv

  24. [32]

    Equivariant Self-Supervision for Musical Tempo Es- timation,

    Elio Quinton, “Equivariant Self-Supervision for Musical Tempo Es- timation,” in International Society for Music Information Retrieval Conference (ISMIR). Sept. 2022, number arXiv:2209.01478, arXiv

    2022 arXiv

  25. [33]

    Towards proper contrastive self-supervised learning strategies for music audio represen- tation,

    Jeong Choi, Seongwon Jang, Hyunsouk Cho, et al., “Towards proper contrastive self-supervised learning strategies for music audio represen- tation,” in 2022 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 2022, pp. 1–6. Appendix for the paper “Leave-One-EquiV...

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.