Pith. sign in

REVIEW 6 major objections 4 minor 40 references

Identification and Clustering of Unseen Ragas in Indian Art Music

T0 review · 6 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper proposes a two-stage pipeline that uses MC-dropout uncertainty to detect out-of-distribution raga audio and contrastive Novel Class Discovery to cluster those unseen ragas, reporting clustering accuracy of 79.34% on PIM and…

desk verdict A serious NCD-for-ragas paper with a real new application, but the clustering evaluation doesn't fully rule out recording-identity confounds, so the semantic claim needs a stronger test. read the letter →

arxiv 2411.18611 v2 pith:VGIPZCVV submitted 2024-11-27 eess.AS

Parampreet Singh , Adwik Gupta , Aakarsh Mishra , Vipul Arora This is my paper · ORCID
classification eess.AS
keywords ragaidentificationopen-setrecognitionout-of-distributiondetectionnovelclassdiscoverycontrastivelearningMonteCarlodropoutIndianartmusicinformationretrieval
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

This paper argues that raga identification should be treated as an open-set problem, because new ragas can appear at test time. It proposes a two-stage pipeline: first, Monte Carlo dropout uncertainty flags out-of-distribution audio clips; second, a Novel Class Discovery stage clusters those clips into distinct unseen raga classes using contrastive learning on unlabeled data. With a CNN-LSTM feature extractor trained on 12 labeled PIM ragas, k-means on the discovered embeddings reaches 79.34% clustering accuracy on PIM and 81.04% on Saraga, close to the clustering quality of supervised methods. The significance is that this would allow music identification systems to use vast unlabeled recordings without expert annotation.

What carries the argument

The carrying mechanism is the combined training objective for the encoder $g(\cdot)$, with pseudo-labeled BCE loss $\ell_{bce}$, hard-negative contrastive loss $\ell_{cl}$, and consistency MSE loss $\ell_{mse}$. Pairwise pseudo-labels $t_{i,j}$ are obtained by thresholding the cosine similarity $\varepsilon$ of $f_{feat}$ embeddings (Eq. 3), so the labeled structure of known ragas is transferred to unlabeled data. Hard negatives are the $H$ least similar embeddings, and the consistency loss ties a clip to time-shifted and volume-modified versions of itself. The openness of the NCD problem is quantified by $O_{NCD} = 1 - \sqrt{2|C_{train}| / (2|C_{train}| + |C_{test}|)}$, and the ablation shows each loss component contributes to the final clustering quality.

What would settle it

An experiment that would settle it: train the same pipeline but replace $f_{feat}$ with a feature extractor trained on scrambled or non-rag audio, then cluster the same five unseen ragas; if clustering accuracy stays near chance, the transfer premise is doing the work. More directly, compute the cosine similarity of $f_{feat}$ embeddings for pairs of clips from the same unseen raga and from different unseen ragas; the method requires that a threshold $\delta$ separates these two distributions, so measuring the overlap of those two distributions would show whether the premise holds.

Watch

Extended reading notes

Core claim

The paper's central claim is that a representation trained only on known ragas carries enough melodic structure that entirely unseen raga classes can be recovered by clustering. The authors train the feature extractor $f_{feat}$ on 12 PIM ragas, then train a self-attention encoder $g$ on unlabeled OOD samples with the combined loss $\ell = \ell_{bce} + \beta\ell_{cl} + \gamma\ell_{mse}$, where cosine-similarity-based pseudo-labels provide BCE supervision, hard negatives provide contrastive supervision, and time/volume shifts provide consistency supervision. The resulting embeddings cluster the five unseen ragas in each dataset with an ACC of 79.34% (PIM) and 81.04% (Saraga) using k-means, outperforming baselines that cluster the raw features directly. The paper also reports that the approach generalizes across datasets, with features trained on PIM clustering Saraga audio successfully.

Load-bearing premise

The load-bearing premise is that the CNN-LSTM feature extractor trained only on the 12 known PIM ragas produces embeddings in which entirely unseen raga classes are already separated by cosine similarity; if that transfer fails, both the pseudo-labels and the final clusters degrade.

Editorial extensions

If this is right

  • If the central claim is correct, unseen ragas can be grouped from unlabeled audio at roughly 79–81% clustering accuracy, reducing the need for expert-labeled data.
  • OOD detection accuracy of 80.87% (PIM) and 85.6% (Saraga) implies that known and unknown ragas can be separated without retraining the classifier.
  • The combined loss beats each component alone (ACC 79.34% vs 76.04% for $\ell_{cl+bce}$, 75.43% for $\ell_{bce}$, 70.16% for $\ell_{cl}$), so the interaction of the three terms is doing real work.
  • Higher openness degrades performance: moving from $O_{NCD}=0.09$ to $0.18$ drops ACC from 79.34% to 55.68%, so the method's usefulness is bounded by how many unseen classes coexist.

Reading between the lines

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

  • Beyond the paper, the pseudo-label threshold $\delta$ is a free parameter that could be tuned per-dataset or per-desired cluster granularity, since no principled method for setting it is proposed.
  • Beyond the paper, the pipeline should be most reliable when unseen ragas are melodically close to the training ragas; a targeted test that varies the thaat distance between seen and unseen ragas would map where transfer breaks.
  • Beyond the paper, the same OOD-plus-contrastive scheme could be applied to other open-set music discovery tasks such as tala or instrument family discovery, since the method is feature-agnostic.
  • Beyond the paper, the similarity-based pseudo-labels may be the main source of confusion for Mishra ragas, which mix notes from multiple parent ragas and produce the lowest per-class F1 scores in the confusion matrix.
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

6 major / 4 minor

Summary. The paper proposes a two-stage pipeline for open-set raga identification: MC-dropout uncertainty is used to detect out-of-distribution audio clips, and a contrastive-learning-based Novel Class Discovery (NCD) module clusters the OOD clips into unseen raga classes. A CNN-LSTM feature extractor f(·) is supervised on 12 known PIM ragas; its penultimate embeddings ffeat(·) are then used to generate pairwise cosine-similarity pseudo-labels (Eqs. 2–3) that train a self-attention encoder g(·) under a combined loss of BCE, contrastive, and consistency terms (Eq. 10). The reported results include OOD accuracy of 85.6% (Saraga) and 80.87% (PIM), and K-means clustering ACC of 79.34% (PIM) and 81.04% (Saraga) for the proposed method versus 70.75% and 75.44% for the baseline (Table 4). The paper also provides loss-component ablations, an openness study, and releases code and metadata.

Significance. If the main claim holds, the paper would make a useful contribution to MIR: it formulates unseen-raga discovery as NCD, defines an openness measure adapted to the problem, and demonstrates a fully unsupervised clustering pipeline that improves over direct clustering of the pretrained embeddings. The release of code and the inclusion of cross-dataset experiments on PIM and Saraga are concrete strengths. However, the significance is tempered by validation gaps: the feature ablation that motivates ffeat(·) compares it against features that were not trained on the same labeled set, the pseudo-label generation is partly self-referential, no error bars or statistical tests are reported, and the paper itself provides evidence that recording/artist identity may drive the Saraga Shree/Marwa confusion. These issues affect the interpretation of the central clustering claim, so the contribution is promising but not yet fully established.

major comments (6)
  1. [Section 3.3.1 and Section 4.2, Eqs. (2)–(3)] The load-bearing premise is that ffeat(·), supervised only on the 12 PIM training ragas, produces embeddings in which entirely unseen raga classes are already separable by cosine similarity. The feature ablation in Table 3 does not establish this premise: ffeat(·) is evaluated on the same labeled set Sl on which it was supervised, reaching ACC 90.05, whereas MERT and melody features were never trained on Sl. A fair transfer test would cluster ffeat(·) embeddings on Su (the unseen raga set) and compare them with frozen general-purpose features on the same Su. As written, the choice of ffeat(·) is supported by a biased comparison, and the pseudo-labels in Eq. (3) inherit whatever structure ffeat(·) has on unseen classes.
  2. [Section 3.3 and Section 4.3] There is no end-to-end evaluation of the complete OOD+NCD pipeline. OOD detection is reported separately in Table 2, and NCD in Table 4 is evaluated on a Su that is assumed to consist entirely of OOD samples. In a real deployment, OOD detection errors would propagate into the clustering stage. I ask for a combined evaluation that first runs MC-dropout OOD detection on a mix of known and unknown classes and then clusters only the samples flagged as OOD, reporting clustering accuracy over the true unseen-class subset and the overall pipeline accuracy.
  3. [Section 4.3.1, Saraga discussion] The paper states that Raag Shree and Raag Marwa are clustered together partly because the recordings feature the same singers and come from the same concert. This is direct evidence that the embeddings may be capturing recording/artist identity rather than raga semantics. Because each Saraga unseen class is represented by at most 3 audio files, a small number of artists or sessions could drive the reported 81.04% ACC. The authors should control for this confound by, for example, splitting by artist/session, reporting per-artist clustering performance, or removing same-concert recordings and re-running Table 4.
  4. [Section 3.5 and Table 4] The ACC evaluation rule invalidates the result if one predicted cluster is matched to multiple true clusters, and the reported numbers give no indication of how often this occurred or how many clusters were discarded. A one-to-one matching rule can make ACC optimistically high relative to a full contingency-table matching. The authors should report the number of valid runs, use a standard Hungarian matching over the full confusion matrix, or at least report ARI as a primary label-dependent metric alongside ACC.
  5. [Section 3.2, Section 3.3.1, and Section 4.1] The OOD variance threshold and the similarity threshold δ are effectively free parameters, yet no sensitivity analysis is provided and the text does not say how the thresholds were chosen. If they are tuned on the very test sets used in Tables 2 and 4, the reported numbers are optimistic. Please report threshold values, show performance across a range of thresholds, or describe a validation-based selection procedure that does not use the final test classes.
  6. [Section 4.3.1, final paragraph] The claim that the proposed clusters 'can approach or match the clustering quality of supervised methods' is not supported by any experiment: Table 4 compares only against unsupervised baselines on Su, and no supervised upper bound for the same unseen classes is reported. Adding a supervised clustering result (for example, k-means on embeddings from a model fine-tuned with labels on Su, or a linear-probe accuracy) would make this claim testable. Until then, the statement should be removed or substantially weakened.
minor comments (4)
  1. [Section 3.3.3, Eq. (9)] The notation is inconsistent: the text defines hard negatives as ξh, but Eq. (9) uses ξm, and Algorithm 1 line 16 uses β for the set of positive embeddings while β is also a loss weight in Eq. (10). Please unify these symbols.
  2. [Section 3.3.2, Eq. (6)] The notation Cl and Cu appears to denote both class counts and class sets; the summation limits should be |Cl| and |Cu| or the sets themselves, and the sums appear to run over samples rather than classes, so the expression as written is unclear.
  3. [Section 3.5, ACC definition] The definition of cpt is ambiguous: it should state clearly that the intersection is over the sets of embeddings assigned to the predicted cluster and the true cluster, and explain how the 'highest overlap' match is computed.
  4. [Section 4.1] The OOD detection section does not report the variance threshold value, the number of threshold trials, or a confusion matrix for the OOD decision; adding these would improve reproducibility.

Circularity Check

2 steps flagged · score 4.0 of 10

Partially self-referential: ffeat feature selection is a training-set evaluation, and BCE pseudo-labels train g to reproduce ffeat's own cosine similarities; the main NCD evaluation remains externally grounded.

  1. fitted input called prediction [Section 4.2, Table 3]
    "The results indicate that embeddings from both MERT and melody-based models yield subpar performance, even when evaluated with label-independent metrics. In contrast, ffeat(·) provides significantly better clustering results. So, we adopt ffeat(·) as the feature extractor for the remainder of our study."

    ffeat(·) is the penultimate layer of f(·), which was trained on Sl with categorical cross-entropy (Section 3.1). Clustering Sl with these embeddings is therefore a training-set evaluation; the 90.05% ACC is a direct consequence of fitting to the same 12 labels, whereas MERT and melody embeddings were not trained on Sl. The paper's conclusion 'So, we adopt ffeat(·)' is thus forced by the experimental construction, not by evidence about transfer to unseen ragas.

  2. self definitional [Section 3.3.1, Eqs. (2)-(5)]
    "Now, using this, we assign a pairwise pseudo-label ti,j as: ti,j = 1[ε(yui, yuj) ≥ δ] ... These pairwise pseudo-labels are used to train a self-attention encoder model g(·) ... The BCE loss function is defined as: ℓbce = ti,j log(pi,j) + (1 − ti,j) log(1 − pi,j)."

    Equation (3) defines t_{i,j} by thresholding cosine similarity in ffeat space, and Eq. (5) trains g so that its output cosine similarity p_{i,j} matches t_{i,j}. Consequently the z embeddings that are subsequently clustered are trained to reproduce the very pairwise similarities of the input embeddings y. The clusters are therefore inherited from ffeat's similarity structure by construction. The external ground-truth ACC evaluation tempers this, so the central claim is not wholly circular.

full rationale

The paper's central NCD pipeline is evaluated against ground-truth labels of unseen raga classes, which is an external benchmark: the reported ACC/ARI are not computed from the same pseudo-labels used to train g, so the main claim retains independent content. There is no load-bearing self-citation chain or uniqueness theorem imported from the authors. However, two steps carry a genuine self-referential burden. First, the feature-extractor selection (Section 4.2) compares ffeat, which was supervised on Sl, with MERT and melody features that were never trained on Sl, so the conclusion that ffeat is superior is an artifact of evaluating on the training set. Second, the BCE training signal in Section 3.3.1 defines pseudo-labels from ffeat's own cosine similarities and then trains g to match those same similarities, meaning the learned z-space clusters are substantially determined by the y-space similarity structure by construction. These issues are partial: the contrastive and consistency losses add independent constraints, the Saraga/PIM unseen classes are disjoint from the labeled classes, and the final evaluation uses true raga labels. The Saraga Shree/Marwa discussion also flags a possible singer/concert confound, which is a validity risk rather than a formal circularity. Overall, the derivation is not equivalent to its inputs, but it is not fully independent of them either.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The pipeline rests on several unverified domain assumptions, most importantly the transferability of a classifier trained on 12 PIM ragas to unseen raga classes. It also contains at least six undisclosed hyperparameters (thresholds, temperatures, loss weights, negative counts) that the central results depend on. No new entities are introduced.

free parameters (7)
  • OOD variance threshold = not stated
    Used to classify each sample as in-distribution or OOD after T=50 MC-dropout passes; no search range or selection criterion is reported (Section 4.1).
  • Similarity threshold δ = not stated
    Eq. (3) turns cosine similarity of ffeat embeddings into pairwise pseudo-labels, the main supervision for the BCE loss; its value is never given.
  • Temperature τ = not stated
    Controls concentration in contrastive loss Eq. (9), inherited from NCL.
  • Loss weights β and γ = not stated
    Scale the contrastive and consistency losses in Eq. (10); authors say proper tuning is critical but give no values.
  • Hard negatives count H = not stated
    Number of least-similar embeddings selected as negatives in Eq. (8).
  • Number of clusters K = 5 (matches the true number of unseen classes)
    K-means and UMAP+K-means assume K equals the true number of unseen classes, which is known in this benchmark; a practical deployment would need to estimate K.
  • MC-dropout passes T = 50
    Number of stochastic forward passes for OOD variance; stated, but the resulting threshold is not.
assumptions (5)
  • domain assumption Tonic-normalized chromagrams capture raga identity
    All inputs are chroma features per [6]; if chroma collapses raga-distinctive melodic information, the entire pipeline is undermined. Section 3.
  • domain assumption A pre-trained CNN-LSTM on 12 PIM ragas transfers to unseen ragas
    The feature extractor ffeat(·) is used both for pseudo-label generation (Eq. 3) and baseline clustering; the paper validates this only indirectly and with a biased ablation. Section 3.3.1, Section 4.2.
  • domain assumption Clips split from the same source audio file share a raga label
    The same-source rule creates positive pairs and BCE pseudo-labels; if a long recording contains multiple ragas or large variations, this assumption weakens. Section 3.3.1.
  • domain assumption The unlabeled OOD set is composed of distinct unseen raga classes with no known-class contamination
    NCD assumes each OOD sample belongs to one of the target novel classes; in this benchmark the test sets are curated to satisfy this. Section 1, 4.1.
  • standard math Uncertainty from MC-dropout is a valid OOD signal
    Borrowed from [14]; accepted in the literature and used without modification. Section 3.2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Identification and Clustering of Unseen Ragas in Indian Art Music." pith.science (2026). https://pith.science/paper/VGIPZCVV

@misc{pith2026241118611,
  author       = {Pith},
  title        = {Pith review of: Identification and Clustering of Unseen Ragas in Indian Art Music},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VGIPZCVV}},
  note         = {Machine review of arXiv:2411.18611}
}
read the original abstract

Raga classification in Indian Art Music is an open-set problem where unseen classes may appear during testing. However, traditional approaches often treat it as a closed set problem, rejecting the possibility of encountering unseen classes. In this work, we try to tackle this problem by first employing an Uncertainty-based Out-Of-Distribution (OOD) detection, given a set containing known and unknown classes. Next, for the audio samples identified as OOD, we employ Novel Class Discovery (NCD) approach to cluster them into distinct unseen Raga classes. We achieve this by harnessing information from labelled data and further applying contrastive learning on unlabelled data. With thorough analysis, we demonstrate the influence of different components of the loss function on clustering performance and examine how varying openness affects the NCD task in hand.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 38 canonical work pages

  1. [1]

    Identi- fication and Clustering of Unseen Ragas in Indian Art Music

    INTRODUCTION Ragas form the core melodic framework of Indian Art Mu- sic (IAM), each characterized by a distinct set of notes and improvisational rules that evoke specific emotions or moods [1]. Identifying Ragas in audio recordings has various applications, including music recommendation systems, cultural preservation, and music education [1]. While trad...

    work page 2025

  2. [2]

    Identification and Clustering of Unseen Ragas in Indian Art Music

    RELATED WORKS 2.1 OOD Detection Uncertainty estimation is a well-established field in ma- chine learning that focuses on evaluating the confidence of model predictions for given test examples. Various ap- proaches utilize uncertainty for identifying OOD samples. The work [9] proposes using maximum softmax probabil- ities as uncertainty indicators. Deep en...

    work page Pith review arXiv 2025

  3. [3]

    We construct a labeled subset S l containing N num- ber of 30-second audio clips xl i, sourced from the PIM dataset [6], each belonging to one of thect predefined Raga Figure 1

    METHOD The overall flow of the whole process is shown in Fig- ure 1. We construct a labeled subset S l containing N num- ber of 30-second audio clips xl i, sourced from the PIM dataset [6], each belonging to one of thect predefined Raga Figure 1. Block diagram illustrating the overall system workflow: audio input is first converted to a chromagram and pro...

    work page

  4. [4]

    EXPERIMENTAL RESULTS The labeled dataset Sl consists of 141 audio files sourced from PIM [6] dataset, segmented into 5,734 audio sam- ples, with a total duration of approximately 47.78 hours. A CNN-LSTM model f (·) is trained in a supervised man- ner on this dataset for multi-class classification across 12 Raga classes, achieving an F1-score of 0.89 throu...

    work page

  5. [5]

    We first use Uncertainty Estimation for Out-of- Distribution (OOD) detection on both the Saraga and PIM datasets, effectively distinguishing unknown Ragas from known ones

    CONCLUSION AND FUTURE SCOPE In this study, we propose a novel approach for identi- fying and clustering unseen Raga classes in Indian Art Music. We first use Uncertainty Estimation for Out-of- Distribution (OOD) detection on both the Saraga and PIM datasets, effectively distinguishing unknown Ragas from known ones. Then, we apply a contrastive learning-ba...

    work page

  6. [6]

    Explainable deep learning analysis for raga identification in indian art music,

    P. Singh and V . Arora, “Explainable deep learning analysis for raga identification in indian art music,”

    work page

  7. [7]

    The computational study of a musical cul- ture through its digital traces,

    X. Serra, “The computational study of a musical cul- ture through its digital traces,” Acta Musicologica , vol. 89, no. 1, p. 24–44, Jun. 2017

    work page 2017

  8. [8]

    Phononet: multi-stage deep neural net- works for raga identification in hindustani classical music,

    S. Chowdhuri, “Phononet: multi-stage deep neural net- works for raga identification in hindustani classical music,” in ICMR, 2019

    work page 2019

Show all 40 references
  1. [9]

    Multimodal Deep Learning Architecture for Hindustani Raga Classifica- tion,

    S. Paschalidou and I. Miliaresi, “Multimodal Deep Learning Architecture for Hindustani Raga Classifica- tion,” Sensors & Transducers, vol. 261, no. 2, pp. 77– 86, Feb. 2024

    2024

  2. [10]

    A north indian raga recognition using ensemble clas- sifier,

    A. A. Bidkar, R. S. Deshpande, and Y . H. Dandawate, “A north indian raga recognition using ensemble clas- sifier,” IJEET, vol. 12, no. 6, pp. 251–258, 2021

    2021

  3. [11]

    Deepsrgm - sequence classification and ranking in indian classical music via deep learning,

    S. T. Madhusudhan and G. V . Chowdhary, “Deepsrgm - sequence classification and ranking in indian classical music via deep learning,” ArXiv, vol. abs/2402.10168, 2024. [Online]. Available: https: //api.semanticscholar.org/CorpusID:208334841

    2024 arXiv

  4. [12]

    Confidence estimation via auxiliary models,

    C. Corbiere, N. Thome, A. Saporta, T.-H. Vu, M. Cord, and P. Perez, “Confidence estimation via auxiliary models,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 10, pp. 6043–6055, 2021

    2021

  5. [13]

    Confidence- enhanced models for indian art music analysis,

    S. Kumar, P. Singh, and V . Arora, “Confidence- enhanced models for indian art music analysis,” in 2025 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), 2025

    2025

  6. [14]

    Toward open set recognition,

    W. J. Scheirer, A. de Rezende Rocha, A. Sapkota, and T. E. Boult, “Toward open set recognition,” IEEE Transactions on Pattern Analysis and Machine Intelli- gence, vol. 35, no. 7, pp. 1757–1772, 2013

    2013

  7. [15]

    Saraga: Open datasets for research on in- dian art music,

    A. Srinivasamurthy, S. Gulati, R. Caro Repetto, and X. Serra, “Saraga: Open datasets for research on in- dian art music,”Empirical Musicology Review, vol. 16, no. 1, p. 85–98, Dec. 2021

    2021

  8. [16]

    A baseline for detecting misclassified and out-of-distribution examples in neu- ral networks,

    D. Hendrycks and K. Gimpel, “A baseline for detecting misclassified and out-of-distribution examples in neu- ral networks,” arXiv preprint arXiv:1610.02136, 2016

    2016 arXiv

  9. [17]

    Simple and scalable predictive uncertainty estimation using deep ensembles,

    B. Lakshminarayanan, A. Pritzel, and C. Blundell, “Simple and scalable predictive uncertainty estimation using deep ensembles,” Advances in neural informa- tion processing systems, vol. 30, 2017

    2017

  10. [18]

    Addressing failure prediction by learning model confidence,

    C. Corbière, N. Thome, A. Bar-Hen, M. Cord, and P. Pérez, “Addressing failure prediction by learning model confidence,” Advances in Neural Information Processing Systems, vol. 32, 2019

    2019

  11. [19]

    Multi-class classification without multi-class labels,

    Y .-C. Hsu, Z. Lv, J. Schlosser, P. Odom, and Z. Kira, “Multi-class classification without multi-class labels,” in International Conference on Learning Representations, 2019. [Online]. Available: https: //openreview.net/forum?id=SJzR2iRcK7

    2019

  12. [20]

    A survey on deep semi-supervised learning,

    X. Yang, Z. Song, I. King, and Z. Xu, “A survey on deep semi-supervised learning,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 9, pp. 8934–8954, 2023

    2023

  13. [21]

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning,

    Y . Gal and Z. Ghahramani, “Dropout as a bayesian approximation: Representing model uncertainty in deep learning,” ininternational conference on machine learning. PMLR, 2016, pp. 1050–1059

    2016

  14. [22]

    Object-graphs for context- aware category discovery,

    Y . J. Lee and K. Grauman, “Object-graphs for context- aware category discovery,” in 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010, pp. 1–8

    2010

  15. [23]

    Automatically discovering and learn- ing new visual categories with ranking statistics,

    K. Han, S.-A. Rebuffi, S. Ehrhardt, A. Vedaldi, and A. Zisserman, “Automatically discovering and learn- ing new visual categories with ranking statistics,” inIn- ternational Conference on Learning Representations , 2020

    2020

  16. [24]

    techniques on spectrogram patches improve instru- ment classification and pitch estimation, though this ap- proach is computationally intensive. [25] utilizes self- supervised contrastive learning for singing voice analysis by applying audio-specific transformations such as ti...

  17. [25]

    Neighborhood contrastive learning for novel class discovery,

    Z. Zhong, E. Fini, S. Roy, Z. Luo, E. Ricci, and N. Sebe, “Neighborhood contrastive learning for novel class discovery,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 10 862–10 870

    2021

  18. [26]

    Learning to discover novel visual categories via deep transfer clustering,

    K. Han, A. Vedaldi, and A. Zisserman, “Learning to discover novel visual categories via deep transfer clustering,” 2019 IEEE/CVF International Conference on Computer Vision (ICCV) , pp. 8400–8408, 2019. [Online]. Available: https://api.semanticscholar.org/ CorpusID:201646290

    2019

  19. [27]

    Unsupervised contrastive learning of sound event representations,

    E. Fonseca, D. Ortego, K. McGuinness, N. E. O’Connor, and X. Serra, “Unsupervised contrastive learning of sound event representations,” in ICASSP 2021 - 2021 IEEE International Conference on Acous- tics, Speech and Signal Processing (ICASSP) , 2021, pp. 371–375

    2021

  20. [28]

    Robust semi-supervised learning by wisely leveraging open- set data,

    Y . Yang, N. Jiang, Y . Xu, and D.-C. Zhan, “Robust semi-supervised learning by wisely leveraging open- set data,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1–15, 2024

    2024

  21. [29]

    Zero-shot learning — the good, the bad and the ugly,

    Y . Xian, B. Schiele, and Z. Akata, “Zero-shot learning — the good, the bad and the ugly,” in 2017 IEEE Con- ference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 3077–3086

    2017

  22. [30]

    Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly,

    Y . Xian, C. H. Lampert, B. Schiele, and Z. Akata, “Zero-shot learning—a comprehensive evaluation of the good, the bad and the ugly,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 41, no. 9, pp. 2251–2265, 2019

    2019

  23. [31]

    Self-supervised learning of audio representations from permutations with differentiable ranking,

    A. N. Carr, Q. Berthet, M. Blondel, O. Teboul, and N. Zeghidour, “Self-supervised learning of audio representations from permutations with differentiable ranking,” IEEE Signal Processing Letters , vol. 28, pp. 708–712, 2021

    2021

  24. [32]

    Self- supervised contrastive learning for singing voices,

    H. Yakura, K. Watanabe, and M. Goto, “Self- supervised contrastive learning for singing voices,” IEEE/ACM Transactions on Audio, Speech, and Lan- guage Processing, vol. 30, pp. 1614–1623, 2022

    2022

  25. [33]

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

    H. Zhao, C. Zhang, B. Zhu, Z. Ma, and K. Zhang, “S3t: Self-supervised pre-training with swin transformer for music classification,” in ICASSP 2022 - 2022 IEEE In- ternational Conference on Acoustics, Speech and Sig- nal Processing (ICASSP), 2022, pp. 606–610

    2022

  26. [35]

    Silhouettes: A graphical aid to the in- terpretation and validation of cluster analysis,

    P. J. Rousseeuw, “Silhouettes: A graphical aid to the in- terpretation and validation of cluster analysis,”Journal of Computational and Applied Mathematics , vol. 20, pp. 53–65, 1987

    1987

  27. [36]

    Information the- oretic measures for clusterings comparison: Variants, properties, normalization and correction for chance,

    N. X. Vinh, J. Epps, and J. Bailey, “Information the- oretic measures for clusterings comparison: Variants, properties, normalization and correction for chance,” Journal of Machine Learning Research , vol. 11, no. 95, pp. 2837–2854, 2010. [Online]. Available: http://jmlr.org/p...

    2010

  28. [37]

    Cluster ensembles - a knowl- edge reuse framework for combining multiple parti- tions,

    A. Strehl and J. Ghosh, “Cluster ensembles - a knowl- edge reuse framework for combining multiple parti- tions,” Journal of Machine Learning Research , vol. 3, pp. 583–617, 01 2002

    2002

  29. [38]

    Interactive singing melody extraction based on active adaptation,

    K. R. Saxena and V . Arora, “Interactive singing melody extraction based on active adaptation,” IEEE/ACM Transactions on Audio, Speech, and Language Pro- cessing, vol. 32, pp. 2729–2738, 2024

    2024

  30. [39]

    MERT: Acoustic music understanding model with large-scale self-supervised training,

    Y . LI, R. Yuan, G. Zhang, Y . Ma, X. Chen, H. Yin, C. Xiao, C. Lin, A. Ragni, E. Benetos, N. Gyenge, R. Dannenberg, R. Liu, W. Chen, G. Xia, Y . Shi, W. Huang, Z. Wang, Y . Guo, and J. Fu, “MERT: Acoustic music understanding model with large-scale self-supervised training,” i...

    2024

  31. [40]

    Understanding the behaviour of contrastive loss,

    F. Wang and H. Liu, “Understanding the behaviour of contrastive loss,” inProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , 2021, pp. 2495–2504

    2021

  32. [2024]

    Available: https://arxiv.org/abs/2406

    [Online]. Available: https://arxiv.org/abs/2406. 02443

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.