Pith. sign in

REVIEW 3 major objections 4 minor 39 references

Multiple Choice Learning for Efficient Speech Separation with Many Speakers

T0 review · 3 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 A quadratic-cost winner-takes-all loss matches PIT for separating up to 20 speakers.

desk verdict Solid empirical extension of MCL to many-speaker separation, but the 'computationally advantageous' claim is overstated for the regime actually studied. read the letter →

arxiv 2411.18497 v1 pith:QCKEVDNE submitted 2024-11-27 cs.SD cs.LGeess.ASstat.ML

David Perera , François Derrida , Théo Mariotte , Gaël Richard , Slim Essid This is my paper · ORCID
classification cs.SDcs.LGeess.ASstat.ML
keywords speechseparationmultiplechoicelearningpermutationinvarianttrainingwinner-takes-allmanyspeakersSI-SDRAUC-SDRcocktailpartyproblem
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 Multiple Choice Learning (MCL), a training objective that assigns each ground-truth speaker to the model's closest prediction and averages those per-target losses, can replace Permutation Invariant Training (PIT) in supervised speech separation. The claim is tested on mixtures of 2 to 5 speakers (WSJ0-mix) and 10 to 20 speakers (LibriMix), where MCL reports SI-SDR scores within a few tenths of a decibel of PIT and SinkPIT, plus similar separation consistency under a new AUC-SDR metric. Because MCL computes its loss in O($n^{2}$) time instead of PIT's O($n^{3}$) optimal-permutation search, the paper argues it is a computationally advantageous substitute as the number of speakers grows. A sympathetic reader would care because many-speaker separation is the regime where PIT's matching cost becomes a bottleneck, and MCL's structure also opens routes to variable-number and unsupervised separation.

What carries the argument

The load-bearing mechanism is the MCL objective (Eq. 4), an averaged target-wise winner-takes-all loss: each ground-truth source $y_i$ is matched to the closest prediction $\hat{y}_j$, and only that winning prediction receives a gradient step on the SI-SDR-based loss $\ell(y_i,\hat{y}_j)$. This is a gradient-descent analogue of K-means and can be computed in $O(n^2)$ time, whereas PIT with the Hungarian algorithm costs $O(n^3)$ and SinkPIT costs $O(n^2/\varepsilon)$ with an approximation controlled by $\varepsilon$. The paper also introduces AUC-SDR, which sorts per-pair SI-SDR scores, normalizes them to $[0,1]$, and averages them, to measure whether separation quality is consistent across all speakers rather than concentrated on a few.

What would settle it

Retrain MCL, PIT, and SinkPIT on WSJ0-mix 5-speaker or LibriMix 10-speaker with several random seeds (e.g., five) and compare SI-SDR means and spreads; if MCL's mean is more than one run-to-run standard deviation below PIT's, the paper's equivalence claim is falsified. A second check: measure loss-computation time at n=50 and n=100 speakers to confirm that MCL's O($n^{2}$) advantage over PIT's O($n^{3}$) materializes in wall-clock time at sizes beyond the paper's 20-speaker experiments.

Watch

Extended reading notes

Core claim

In the paper's own terms, the central discovery is that MCL matches the performances of PIT while being computationally advantageous. Concretely, the paper reports SI-SDR values on the test sets of WSJ0-mix 2/3/4/5 speakers and LibriMix 10/20 speakers in which MCL stays within roughly 0.2 dB of Hungarian-PIT and within about 0.3 dB of SinkPIT, with the largest gap occurring at 5 speakers. It also reports that the average loss-computation time per sample reflects the theoretical complexities as the speaker count rises, and that a new AUC-SDR metric shows comparable separation consistency across sources for all three objectives. The paper interprets these results as evidence that MCL is a viable alternative to optimal-permutation training for many-speaker separation, despite lacking a guarantee of finding the optimal prediction-target matching.

Load-bearing premise

The claim that MCL 'matches' PIT rests on a single training run per method and dataset, with no standard deviations or significance tests, so observed gaps up to 0.22 dB could fall inside normal run-to-run variability.

Editorial extensions

If this is right

  • MCL can serve as a drop-in training objective for many-speaker separation models, removing the need for Hungarian or Sinkhorn matching inside the loss.
  • For speaker counts at which PIT's optimal-permutation search dominates training time, MCL's quadratic loss should scale more gracefully; the paper measures this advantage directly in per-sample loss-computation time.
  • MCL's specialization behavior suggests it can be paired with scoring heads to handle a variable number of speakers, and with few-speaker training data to support unsupervised separation, as the paper states.
  • Separation consistency measured by AUC-SDR falls as speaker count grows for all objectives, so MCL inherits the same limitation as PIT: optimal matching losses do not by themselves enforce uniform quality across sources.

Reading between the lines

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

  • If MCL's equivalence holds under repeated training runs, the practical bottleneck for very large speaker counts shifts from the matching objective to the model architecture, and MCL's O(n^2) loss could be tested well beyond 20 speakers where PIT becomes forbiddingly slow.
  • Because MCL is a clustering-like objective, its performance may degrade if hypotheses collapse onto the same target in harder conditions; annealing or stochastic selection, which the paper only mentions as prior work, would be the natural corrective to test.
  • The AUC-SDR metric could be applied beyond permutation-based training, e.g., to compare models on the worst-separated speaker or to shape losses that explicitly penalize low-performer speakers, a direction the paper leaves to future work.
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. This manuscript proposes Multiple Choice Learning (MCL) as a training objective for supervised speech separation with many speakers, replacing PIT's exhaustive or Hungarian matching with an averaged target-wise winner-takes-all loss (Eq. 4). Using the Swave architecture, the authors compare MCL against PIT and SinkPIT on WSJ0-mix (2–5 speakers) and LibriMix (10–20 speakers), reporting SI-SDR in Table II, a newly introduced AUC-SDR consistency metric in Table III, and timing measurements in Figure 2. The paper concludes that MCL matches PIT in separation quality while being computationally advantageous, and suggests extensions to variable-number and unsupervised separation settings.

Significance. If the equivalence claim held with proper uncertainty quantification, the result would be practically useful: MCL has O(n^2) loss evaluation versus O(n^3) for Hungarian PIT, and the experiments cover a wider range of speaker counts (up to 20) than most prior separation work. The manuscript's methodology is mostly sound: it reimplements the baseline losses in a common architecture, uses standard external benchmarks, and includes an explicit timing study rather than relying only on asymptotic arguments. However, the central 'on par and computationally advantageous' claim is currently supported only by single-run point estimates and by a loss-only timing extrapolation that the paper itself acknowledges does not carry over to epoch-level training time in the studied regime. The AUC-SDR metric is an interesting secondary contribution, but its interpretation and invariance properties need more discussion.

major comments (3)
  1. [Abstract; Section IV-B; Figure 2] The abstract and conclusion state that MCL is 'computationally advantageous' relative to PIT, but the paper's own epoch-duration results in Section IV-B (right panel of Figure 2) show that for n ≤ 20, the loss-computation gap is negligible relative to the rest of the training pipeline, and the only visible advantage appears in the isolated loss-time extrapolation for n ≈ 100, which the same section calls 'unlikely for speech separation.' At minimum, the claim should be qualified to asymptotic loss-computation complexity rather than end-to-end training time, and the abstract/conclusion should not present computational advantage as an experimentally established property in the many-speaker settings studied.
  2. [Tables II–III; Section IV-A] The 'on par with PIT' conclusion rests on single-run SI-SDR values with no seeds, standard deviations, or significance tests. The largest differences in Table II (e.g., 0.22 dB against PIT on 2- and 5-speaker WSJ0-mix, and 0.68 dB in favor of MCL on 10-speaker LibriMix) are within the typical run-to-run variation for this type of model, so the equivalence could disappear under repeated training. Please report multiple seeds with means and error bars, or at least a paired significance test, for the central performance comparisons in Tables II and III.
  3. [Section II-D; Section IV-C] The AUC-SDR metric is not yet convincingly established as a measure of consistency. Since AUC-SDR is an affine transformation of the mean sorted SI-SDR with constants set by the sample maximum and minimum, a value near 0.5 does not by itself imply that 'a majority of speakers are well separated' (Section IV-C); it only locates the average between the best and worst pair. The choice of min(0, s_n) as the lower anchor also introduces a dataset-dependent floor. Please clarify the intended interpretation and validate AUC-SDR against explicit failure distributions (e.g., one poorly separated speaker among many) before using it to draw conclusions about consistency.
minor comments (4)
  1. [Section III-A] The text contains a typo: '10 ans 20 speakers' should read '10 and 20 speakers.'
  2. [Section IV-B; Figure 2] The left panel of Figure 2 reports loss-computation times for large speaker counts, but the measurement procedure is not described: number of samples averaged, hardware, implementation details of the Hungarian/Sinkhorn solvers, and whether the timings include the backward pass. Please add these details so the timing comparison is reproducible.
  3. [Section II-B; Equation (3)] The sign and role of the entropy regularization in the SinkPIT objective should be stated more carefully; in particular, the dependence of the approximation error on ε and the direction of the limit ε → 0 should be made explicit.
  4. [Section IV-D] The manuscript asserts that MCL collapse 'is not encountered in practice' but provides no direct diagnostic. Reporting the fraction of hypotheses selected by the min operation in Eq. (4) over a validation set would make this claim checkable, since aggregate SI-SDR alone may be insensitive to a rarely selected degenerate output.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; MCL performance claim is grounded in external benchmarks and fresh reimplementation.

full rationale

The paper's central claim is that the MCL objective (Eq. 4) matches PIT and SinkPIT on WSJ0-mix and LibriMix. These are external benchmarks and the baselines are reimplemented under the same experimental settings, so the comparison is not defined in terms of MCL itself. Eq. (4) is inherited from the authors' prior work [23], but it is restated, reimplemented, and evaluated on new many-speaker setups; the prior result is not used as evidence for the new equivalence, so the self-citation is not load-bearing. No fitted parameter is renamed as a prediction: the reported SI-SDR values are direct evaluation outcomes, and AUC-SDR is a descriptive consistency metric rather than an input to the main derivation. The paper's own Section IV-B acknowledges that the computational advantage is negligible for the tested n <= 20 regime and only appears for n approaching 100, which it calls unlikely for speech separation. This is an internal consistency/correctness tension with the abstract's wording, not a circularity: the complexity analysis is explicit and the performance comparison remains externally grounded. No circular step can be exhibited by reducing an equation or result to its own input.

Assumptions & free parameters 0 free parameters · 6 assumptions · 1 invented entities

No fitted free parameters appear in this paper; architecture hyperparameters are constants inherited from Swave [27]. The only new construct is the AUC-SDR metric, which has no external evidence. All background rests on standard speech-separation assumptions and methods inherited from prior work.

assumptions (6)
  • domain assumption SI-SDR (Eq. 1) is a valid pairwise reconstruction metric for speech separation.
    Standard in the field; cited [13]. Used as both a training loss and the evaluation metric.
  • standard math PIT and SinkPIT objectives (Eqs. 2-3) give the optimal prediction-target assignment when epsilon vanishes.
    Cites Hungarian algorithm and Sinkhorn optimal transport results; serves as the reference baseline.
  • domain assumption Target-wise averaged winner-takes-all loss (Eq. 4) is a suitable objective for speech separation.
    Adopted from the authors' prior work [23]; no derivation or robustness analysis is provided in this paper.
  • domain assumption Swave is a competitive architecture for many-speaker separation.
    Architecture taken from Nachmani et al. [27] and Dovrat et al. [19]; its adequacy for 10-20 speakers is assumed rather than re-verified.
  • domain assumption Optimal permutation SI-SDR is the standard evaluation metric; AUC-SDR is a meaningful complement.
    Optimal permutation SI-SDR is standard in [19,27]. AUC-SDR is newly defined and assumed to capture consistency.
  • ad hoc to paper AUC-SDR, as normalized by mapping the highest score to 1 and min(0, s_n) to 0, measures separation consistency meaningfully.
    The normalization and the empirical-mean construction in Section II-D are chosen for this paper without external validation or comparison to alternative consistency measures.
invented entities (1)
  • AUC-SDR
    purpose: Evaluates the consistency of SI-SDR scores across the n prediction-target pairs, rather than only their mean.
    Defined in Section II-D and Figure 1. It is a descriptive statistic; no external validation, theoretical motivation, or downstream-task correlation is provided, so it lacks independent falsifiable support.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multiple Choice Learning for Efficient Speech Separation with Many Speakers." pith.science (2026). https://pith.science/paper/QCKEVDNE

@misc{pith2026241118497,
  author       = {Pith},
  title        = {Pith review of: Multiple Choice Learning for Efficient Speech Separation with Many Speakers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QCKEVDNE}},
  note         = {Machine review of arXiv:2411.18497}
}
read the original abstract

Training speech separation models in the supervised setting raises a permutation problem: finding the best assignation between the model predictions and the ground truth separated signals. This inherently ambiguous task is customarily solved using Permutation Invariant Training (PIT). In this article, we instead consider using the Multiple Choice Learning (MCL) framework, which was originally introduced to tackle ambiguous tasks. We demonstrate experimentally on the popular WSJ0-mix and LibriMix benchmarks that MCL matches the performances of PIT, while being computationally advantageous. This opens the door to a promising research direction, as MCL can be naturally extended to handle a variable number of speakers, or to tackle speech separation in the unsupervised setting.

Figures

Figures reproduced from arXiv: 2411.18497 by the authors.

Figure 1
Figure 1. Schematic representation of AUC-SDR. No mechanism ensures that all the hypotheses yˆi are selected, an issue known as collapse [29]. In particular, MCL is not guaranteed to find the optimal prediction–target matching. This is unlike PIT and SinkPIT, which therefore act as upper bound references in our experiments. Nonetheless, we demonstrate empirically that this problem is not encountered in practice (see Section I… view at source ↗
Figure 2
Figure 2. Time complexity of MCL, PIT and SinkPIT. On the left, we show the average computation time per sample of MCL (orange dashed line), PIT (blue solid line) and SinkPIT (green dotted line) separation losses, as a function of the number of speakers. On the right, we display the relative training time over one epoch of MCL and SinkPIT, computed for the WSJ0- mix datasets (2 to 5 speakers) and LibriMix datasets (10 and 20 … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 29 canonical work pages

  1. [1]

    Automatic speech recognition in cocktail-party situations: A specific training for separated speech,

    A. Marti, M. Cobos, and J. J. Lopez, “Automatic speech recognition in cocktail-party situations: A specific training for separated speech,” The Journal of the Acoustical Society of America , vol. 131, no. 2, pp. 1529–1535, 2012

    work page 2012

  2. [2]

    Espnet-se: End- to-end speech enhancement and separation toolkit designed for asr integration,

    C. Li, J. Shi, W. Zhang, A. S. Subramanian, X. Chang, N. Kamo, M. Hira, T. Hayashi, C. Boeddeker, Z. Chen et al. , “Espnet-se: End- to-end speech enhancement and separation toolkit designed for asr integration,” in Spoken Language Technology Workshop (SLT) , 2021, pp. 785–792

    work page 2021

  3. [3]

    All-neural online source separation, counting, and diarization for meeting analysis,

    T. V on Neumann, K. Kinoshita, M. Delcroix, S. Araki, T. Nakatani, and R. Haeb-Umbach, “All-neural online source separation, counting, and diarization for meeting analysis,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2019, pp. 91–95

    work page 2019

  4. [4]

    Ts-sep: Joint diarization and separation conditioned on estimated speaker embeddings,

    C. Boeddeker, A. S. Subramanian, G. Wichern, R. Haeb-Umbach, and J. Le Roux, “Ts-sep: Joint diarization and separation conditioned on estimated speaker embeddings,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , 2024

    work page 2024

  5. [5]

    Singing-voice separation from monaural recordings using robust prin- cipal component analysis,

    P.-S. Huang, S. D. Chen, P. Smaragdis, and M. Hasegawa-Johnson, “Singing-voice separation from monaural recordings using robust prin- cipal component analysis,” in IEEE International Conference on Acous- tics, Speech and Signal Processing (ICASSP) , 2012, pp. 57–60

    work page 2012

  6. [6]

    Speech recognition by bilateral cochlear implant users in a cocktail- party setting,

    P. C. Loizou, Y . Hu, R. Litovsky, G. Yu, R. Peters, J. Lake, and P. Roland, “Speech recognition by bilateral cochlear implant users in a cocktail- party setting,” The Journal of the Acoustical Society of America , vol. 125, no. 1, pp. 372–383, 2009

    work page 2009

  7. [7]

    The cocktail party robot: Sound source separation and localisation with an active binaural head,

    A. Deleforge and R. Horaud, “The cocktail party robot: Sound source separation and localisation with an active binaural head,” in Proceedings of the seventh annual ACM/IEEE international conference on Human- Robot Interaction, 2012, pp. 431–438

    work page 2012

  8. [8]

    End- to-end speaker diarization for an unknown number of speakers with encoder-decoder based attractors,

    S. Horiguchi, Y . Fujita, S. Watanabe, Y . Xue, and K. Nagamatsu, “End- to-end speaker diarization for an unknown number of speakers with encoder-decoder based attractors,” arXiv preprint arXiv:2005.09921 , 2020

    arXiv 2005

Show all 39 references
  1. [9]

    Boosting unknown-number speaker separation with transformer decoder-based attractor,

    Y . Lee, S. Choi, B.-Y . Kim, Z.-Q. Wang, and S. Watanabe, “Boosting unknown-number speaker separation with transformer decoder-based attractor,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2024, pp. 446–450

    2024

  2. [10]

    Looking to listen at the cocktail party: a speaker-independent audio-visual model for speech separation,

    A. Ephrat, I. Mosseri, O. Lang, T. Dekel, K. Wilson, A. Hassidim, W. T. Freeman, and M. Rubinstein, “Looking to listen at the cocktail party: a speaker-independent audio-visual model for speech separation,” ACM Trans. Graph. , vol. 37, no. 4, jul 2018. [Online]. Available: htt...

    2018

  3. [11]

    SepIt: Approaching a single channel speech separation bound,

    S. Lutati, E. Nachmani, and L. Wolf, “SepIt: Approaching a single channel speech separation bound,” type: article. [Online]. Available: http://arxiv.org/abs/2205.11801

    arXiv

  4. [12]

    Supervised speech separation based on deep learning: An overview,

    D. Wang and J. Chen, “Supervised speech separation based on deep learning: An overview,” IEEE/ACM transactions on audio, speech, and language processing, vol. 26, no. 10, pp. 1702–1726, 2018

    2018

  5. [13]

    Sdr–half-baked or well done?

    J. Le Roux, S. Wisdom, H. Erdogan, and J. R. Hershey, “Sdr–half-baked or well done?” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2019, pp. 626–630

    2019

  6. [14]

    Joint optimization of masks and deep recurrent neural networks for monaural source separation,

    P.-S. Huang, M. Kim, M. Hasegawa-Johnson, and P. Smaragdis, “Joint optimization of masks and deep recurrent neural networks for monaural source separation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 23, no. 12, pp. 2136–2147, 2015

    2015

  7. [15]

    Deep neural networks for single-channel multi-talker speech recognition,

    C. Weng, D. Yu, M. L. Seltzer, and J. Droppo, “Deep neural networks for single-channel multi-talker speech recognition,” IEEE/ACM Trans- actions on Audio, Speech, and Language Processing , vol. 23, no. 10, pp. 1670–1679, 2015

    2015

  8. [16]

    Deep clustering: Discriminative embeddings for segmentation and separation,

    J. R. Hershey, Z. Chen, J. Le Roux, and S. Watanabe, “Deep clustering: Discriminative embeddings for segmentation and separation,” in IEEE international conference on acoustics, speech and signal processing (ICASSP), 2016, pp. 31–35

    2016

  9. [17]

    Deep attractor network for single- microphone speaker separation,

    Z. Chen, Y . Luo, and N. Mesgarani, “Deep attractor network for single- microphone speaker separation,” in IEEE international conference on acoustics, speech and signal processing (ICASSP) , 2017, pp. 246–250

    2017

  10. [18]

    Permutation invariant training of deep models for speaker-independent multi-talker speech separation,

    D. Yu, M. Kolbæk, Z.-H. Tan, and J. Jensen, “Permutation invariant training of deep models for speaker-independent multi-talker speech separation,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2017, pp. 241–245

    2017

  11. [19]

    Many-speakers single channel speech separation with optimal permutation training,

    S. Dovrat, E. Nachmani, and L. Wolf, “Many-speakers single channel speech separation with optimal permutation training,” in Interspeech 2021, 2021, pp. 3890–3894

    2021

  12. [20]

    Theoretical improvements in algorithmic efficiency for network flow problems,

    J. Edmonds and R. M. Karp, “Theoretical improvements in algorithmic efficiency for network flow problems,” Journal of the ACM (JACM) , vol. 19, no. 2, pp. 248–264, 1972

    1972

  13. [21]

    Towards listening to 10 people simultaneously: An efficient permutation invariant training of audio source separation using sinkhorn’s algorithm,

    H. Tachibana, “Towards listening to 10 people simultaneously: An efficient permutation invariant training of audio source separation using sinkhorn’s algorithm,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2021, pp. 491–495

    2021

  14. [22]

    Sinkhorn distances: Lightspeed computation of optimal transport,

    M. Cuturi, “Sinkhorn distances: Lightspeed computation of optimal transport,” Advances in neural information processing systems , vol. 26, 2013

    2013

  15. [23]

    Annealed multiple choice learning: Overcoming limitations of winner-takes-all with annealing,

    D. Perera, V . Letzelter, T. Mariotte, A. Cort ´es, M. Chen, S. Essid, and G. Richard, “Annealed multiple choice learning: Overcoming limitations of winner-takes-all with annealing,” arXiv preprint arXiv:2407.15580 , 2024

    2024 arXiv

  16. [24]

    Multiple choice learning: Learning to produce multiple structured outputs,

    A. Guzman-Rivera, D. Batra, and P. Kohli, “Multiple choice learning: Learning to produce multiple structured outputs,” Advances in neural information processing systems , vol. 25, 2012

    2012

  17. [25]

    Stochastic multiple choice learning for training diverse deep ensembles,

    S. Lee, S. Purushwalkam Shiva Prakash, M. Cogswell, V . Ranjan, D. Crandall, and D. Batra, “Stochastic multiple choice learning for training diverse deep ensembles,” Advances in Neural Information Processing Systems, vol. 29, 2016

    2016

  18. [26]

    Dsmcl: Dual-level stochas- tic multiple choice learning for multi-modal trajectory prediction,

    Z. Wang, S. Zhou, Y . Huang, and W. Tian, “Dsmcl: Dual-level stochas- tic multiple choice learning for multi-modal trajectory prediction,” in International Conference on Intelligent Transportation Systems (ITSC) . IEEE, 2020, pp. 1–6

    2020

  19. [27]

    V oice separation with an unknown number of multiple speakers,

    E. Nachmani, Y . Adi, and L. Wolf, “V oice separation with an unknown number of multiple speakers,” in Proceedings of the 37th International Conference on Machine Learning . PMLR, 2020, pp. 7164–7175, ISSN: 2640-3498. [Online]. Available: https: //proceedings.mlr.press/v119/na...

    2020

  20. [28]

    Least squares quantization in pcm,

    S. Lloyd, “Least squares quantization in pcm,” IEEE transactions on information theory, vol. 28, no. 2, pp. 129–137, 1982

    1982

  21. [29]

    Learning in an uncertain world: Representing ambiguity through multiple hypotheses,

    C. Rupprecht, I. Laina, R. DiPietro, M. Baust, F. Tombari, N. Navab, and G. D. Hager, “Learning in an uncertain world: Representing ambiguity through multiple hypotheses,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 3591–3600

    2017

  22. [30]

    Librimix: An open-source dataset for generalizable speech separation,

    J. Cosentino, M. Pariente, S. Cornell, A. Deleforge, and E. Vincent, “Librimix: An open-source dataset for generalizable speech separation,” arXiv preprint arXiv:2005.11262 , 2020

    2005 arXiv

  23. [31]

    Conv-tasnet: Surpassing ideal time– frequency magnitude masking for speech separation,

    Y . Luo and N. Mesgarani, “Conv-tasnet: Surpassing ideal time– frequency magnitude masking for speech separation,” IEEE/ACM trans- actions on audio, speech, and language processing , vol. 27, no. 8, pp. 1256–1266, 2019

    2019

  24. [32]

    Dual-path RNN: Efficient long sequence modeling for time-domain single-channel speech separation,

    Y . Luo, Z. Chen, and T. Yoshioka, “Dual-path RNN: Efficient long sequence modeling for time-domain single-channel speech separation,” in International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 46–50, ISSN: 2379-190X. [Online]. Available: https:...

    2020

  25. [33]

    Attention is all you need in speech separation,

    C. Subakan, M. Ravanelli, S. Cornell, M. Bronzi, and J. Zhong, “Attention is all you need in speech separation,” in International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2021, pp. 21–25, ISSN: 2379-190X. [Online]. Available: https: //ieeexplore.ieee.or...

    2021

  26. [34]

    Mossformer: Pushing the performance limit of monaural speech separation using gated single-head transformer with convolution-augmented joint self-attentions,

    S. Zhao and B. Ma, “Mossformer: Pushing the performance limit of monaural speech separation using gated single-head transformer with convolution-augmented joint self-attentions,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2023, pp. 1–5

    2023

  27. [35]

    Tf-gridnet: Integrating full-and sub-band modeling for speech separation,

    Z.-Q. Wang, S. Cornell, S. Choi, Y . Lee, B.-Y . Kim, and S. Watan- abe, “Tf-gridnet: Integrating full-and sub-band modeling for speech separation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2023

    2023

  28. [36]

    Universal sound separation,

    I. Kavalerov, S. Wisdom, H. Erdogan, B. Patton, K. Wilson, J. Le Roux, and J. R. Hershey, “Universal sound separation,” in WASPAA. IEEE, 2019, pp. 175–179

    2019

  29. [37]

    Signal source separation in the analysis of neural activity in brain,

    Y . Zheng, D. Johnston, J. Berwick, and J. Mayhew, “Signal source separation in the analysis of neural activity in brain,” Neuroimage, vol. 13, no. 3, pp. 447–458, 2001

    2001

  30. [38]

    Deep neural network techniques for monaural speech enhancement and separation: state of the art analysis,

    P. Ochieng, “Deep neural network techniques for monaural speech enhancement and separation: state of the art analysis,” Artificial Intelli- gence Review, vol. 56, no. Suppl 3, pp. 3651–3703, 2023

    2023

  31. [39]

    Resilient multiple choice learning: A learned scoring scheme with application to audio scene analysis,

    V . Letzelter, M. Fontaine, M. Chen, P. P ´erez, S. Essid, and G. Richard, “Resilient multiple choice learning: A learned scoring scheme with application to audio scene analysis,” Advances in neural information processing systems, vol. 36, 2024

    2024

Pith tools

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