Pith. sign in

REVIEW 4 major objections 6 minor 45 references

AnalyticKWS: Towards Exemplar-Free Analytic Class Incremental Learning for Small-footprint Keyword Spotting

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a keyword-spotting model can keep learning new spoken commands from a closed-form recursive matrix update, storing no old audio, and reach accuracy matching joint training on all tasks.

desk verdict A useful exemplar-free KWS recipe with strong numbers, but the paper's central proof is botched as printed and needs correction and code before the joint-learning claim can be trusted. read the letter →

arxiv 2505.11817 v1 pith:RXTWKIVJ submitted 2025-05-17 eess.AS cs.LGcs.SD

Yang Xiao , Tianyi Peng , Rohan Kumar Das , Yuchen Hu , Huiping Zhuang This is my paper · ORCID
classification eess.AScs.LGcs.SD
keywords analyticlearningclassincrementalexemplar-freecontinualkeywordspottingcatastrophicforgettingrecursiveleastsquaresacousticfeatureexpansionsmall-footprintKWS
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 show that keyword spotting can continually add new words without ever replaying old audio, by replacing gradient-based fine-tuning with a closed-form recursive least-squares update. If correct, continual learning on small devices no longer requires an exemplar buffer, so user data stays on-device and private. Each new task is adapted in a single epoch, without back-propagation, and the only stored artifact is a small square correlation matrix. The experiments report near-joint accuracy and near-zero backward transfer across up to 100 keywords, with far lower training time and memory than rehearsal-based baselines.

What carries the argument

The load-bearing object is the acoustic feature auto-correlation matrix (AFAM), $A_{t-1} = \left(\sum_{i=0}^{t-1} S_i'^\top S_i' + \gamma I\right)^{-1}$, together with the recursive classifier update asserted in Eq. (7). The AFAM encodes the inverse regularized covariance of all expanded features seen so far; updating it and the classifier via the Woodbury matrix identity reproduces the joint least-squares solution without revisiting old data. A fixed random linear expansion layer, the acoustic feature expansion (AFE), raises the feature dimension before the fit, and the paper's ablation shows accuracy rising with expansion size.

What would settle it

A decisive check is to compare, on the same held-out feature matrices, the weights produced by the recursive update of Eq. (7) with those of the direct joint least-squares solution of Eq. (5): any difference beyond floating-point error would refute the claimed identity. Separately, if task 0 is drawn from one acoustic domain and later tasks from another, and final accuracy falls well short of the joint fit on the same frozen features, the frozen-extractor assumption is the failure point.

Watch

Extended reading notes

Core claim

The central claim is that, once a CNN feature extractor is frozen and the acoustic features are passed through a fixed random linear expansion, the incremental classifier update computed only from current-task features and a stored correlation matrix is algebraically identical to the joint analytic least-squares solution over all tasks. The paper states this in Section 3.2.3: the recursive formulation gives the same answer as joint learning. Consequently, each new keyword task can be learned in one epoch without back-propagation and without storing any past utterances, because the acoustic feature auto-correlation matrix (AFAM) summarizes all past feature correlations and is itself updated recursively. The reported evidence is that AnalyticKWS matches or exceeds rehearsal-based baselines in accuracy and backward transfer on GSC-v1, GSC-v2, and SC-100, while using only a few seconds per task and a tiny constant memory footprint.

Load-bearing premise

The load-bearing premise is that a CNN feature extractor, pretrained only on the first task and then frozen, yields features in which all later, unseen keywords are linearly separable after a fixed random expansion.

Editorial extensions

If this is right

  • No user audio needs to be stored for continual KWS adaptation, so privacy requirements such as user deletion requests are satisfied by construction.
  • Per-task adaptation requires one epoch and no gradient back-propagation, cutting training time on the reported hardware from hundreds of seconds to single digits per task.
  • The AFAM is a square matrix whose size depends only on the expanded feature dimension (e.g., 128 by 128), so extra memory stays small and constant as the number of tasks grows.
  • Accuracy and backward transfer approach those of joint training, with near-zero BWT across up to 51 tasks even when rehearsal baselines use a 500- or 1000-sample buffer.
  • Small-footprint backbones like TC-ResNet-8 can keep incremental keyword spotting on-device, because the only saved artifact is the AFAM plus the current classifier weights.

Reading between the lines

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

  • The recursive identity behind the AFAM is not specific to speech; the same update should transfer to any class-incremental problem with a frozen feature extractor, meaning the privacy and memory benefits are generic properties of analytic continual learning rather than of keyword spotting.
  • Since the frozen task-0 feature extractor is the single point of failure, a natural testable extension is to periodically adapt the expansion layer or maintain a small set of feature anchors from the current task, trading a little privacy for plasticity; the paper's own limitation section points at this gap.
  • The flat accuracy across T=11 to T=51 suggests that, for a fixed frozen backbone, the bottleneck is feature quality rather than forgetting; comparing different backbone capacities under the same analytic update would isolate how much of the gain comes from the representation versus the update rule.
  • The ablation trend of improving accuracy with expansion size up to 512 dimensions suggests a testable extension: whether even larger, or learned rather than random, expansions continue to help or saturate.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes AnalyticKWS, an exemplar-free class-incremental learning method for small-footprint keyword spotting. A CNN backbone is first trained on task 0 and then frozen. A random linear expansion layer (AFE) is inserted, and the classifier is recalibrated with ridge regression. For each subsequent keyword task, the method updates a stored feature autocorrelation matrix (AFAM) and the analytic classifier through a recursive least-squares rule, claiming that this incremental update is identical to the joint analytic least-squares solution over all tasks. Experiments on GSC-v1, GSC-v2, and SC-100 with up to 51 tasks compare against EWC, iCaRL, BiC, Rwalk, RK, and DE-KWS, reporting accuracy, backward transfer, per-epoch training time, and extra memory. The reported results show strong accuracy and much lower training time than the baselines, without storing any exemplars.

Significance. If the recursive equivalence is correctly established and the experimental results are reproducible, the method is genuinely attractive for edge-device KWS: it is exemplar-free, privacy-preserving, requires only one analytic update per task, and reports near-joint accuracy with very low per-task training time. The paper has clear strengths: it tackles a practical problem, evaluates on three datasets including a 100-class setting, reports computational efficiency, and builds on a standard recursive-least-squares formulation rather than ad hoc fitting. However, the central theoretical derivation as printed contains a malformed update equation and an appendix proof with sign and algebraic errors, so the paper's strongest claim — that incremental learning gives the same answer as joint learning — is not currently supported. The frozen-feature premise is also load-bearing and is only partially acknowledged in the limitations paragraph.

major comments (4)
  1. [§3.2.3, Eq. (7)] The recursive update printed as Eq. (7) is not a well-formed matrix expression: it reads θ_t = [θ_{t−1} − A_t S_t^T S'_t θ_{t−1} A_t S'_t^T y_t], with no operator between the last two terms, and it mixes S_t and S'_t notation. If the intended formula is the standard recursive least-squares update, it should take the form θ_t = θ_{t−1} + A_t S_t'^T (Y_t − S_t' θ_{t−1}) (or an algebraically equivalent expression). As printed, Eq. (7) cannot be evaluated and is not shown to equal Eq. (5). This is the central theoretical guarantee of the paper and must be corrected with a full, valid derivation of the exact update implemented in the experiments.
  2. [Appendix A, steps (b) and (d)] The appendix proof contains a sign error and a dropped term. Step (b) writes θ_t = A_t(Q_{t−1} − S_t'^T Y_t), but the joint solution in Eq. (5) requires a plus sign, Q_{t−1} + S_t'^T Y_t. Step (d) then reduces the expression to θ_t = θ_{t−1} − A_t S_t'^T θ_{t−1}, which discards the new-task term Y_t. The final line of the proof is therefore not equivalent to Eq. (5). The recursive-least-squares identity is standard and likely repairable with a correct Woodbury-based derivation, but the proof as written does not establish the claimed equality, and the manuscript cannot be accepted with this defect in the core argument.
  3. [§3.2.1 and Limitations] The method's practical accuracy depends on the premise that features from the task-0-trained, frozen CNN extractor, after a random linear expansion, are linearly separable for all future unseen keywords. The analytic update and AFAM can only rearrange the fixed feature space; they cannot recover information lost because the backbone is never adapted. The limitations paragraph concedes that the extractor 'might not be optimal for every domain.' Because this assumption is load-bearing for the claim of near-joint accuracy, the paper should provide evidence: for example, vary the size and composition of task 0, measure linear separability of the AFE features for future tasks, or show what happens when task 0 is small or unrepresentative.
  4. [Table 1] The Joint upper-bound row reports only three accuracy values (94.93, 94.76, 95.32) for the nine experimental settings; no joint accuracy is reported for GSC-v2 or SC-100. Since the paper repeatedly claims that AnalyticKWS approaches joint-training performance, missing the upper bound for most settings weakens this claim. Please report joint accuracy for every dataset and task configuration, or explicitly state that a single joint model is used for all settings and justify why that is appropriate.
minor comments (6)
  1. [§3.2.3, Eq. (7)] The notation is inconsistent: the left factor uses S_t and the right factor uses S'_t; the expanded feature matrix should be denoted uniformly, e.g., S'_t throughout the update.
  2. [§3.2.2, Eq. (2)] The label matrix y_0 is used without definition; please state explicitly that it is a one-hot encoded matrix whose rows correspond to the samples in S'_0.
  3. [Appendix A] There is a typo in the text: 'AFAuM' should be 'AFAM.' The sentence 'Now we proof calculate' is also ungrammatical.
  4. [§5.3] The memory comparison is hard to follow because the dimensions behind '16000×1×1×500' are not explained. Please give the tensor shape or number of stored scalars for each method.
  5. [Table 3] The ablation finds the best accuracy with a 512-dimensional expansion, but the main results use AnalyticKWS-128 and AnalyticKWS-256. Please clarify why the best configuration is not used in the headline comparisons, or add AnalyticKWS-512 to Table 1.
  6. [§4.2 and Appendix C] The experimental setup does not specify the regularization coefficient γ, the pretraining epochs and optimizer for task 0, or the hyperparameters for baselines such as EWC and BiC. These details should be provided to make the comparisons reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the recursive update is presented as a closed-form least-squares identity, and the paper's self-citations are to externally published prior work, not to the target result.

full rationale

The AnalyticKWS derivation is not circular. The feature extractor is pretrained only on task 0 and then frozen; the analytic classifier in Eq. (3) solves the task-0 ridge regression, Eq. (5) is the closed-form solution of the joint objective Eq. (4), and Eqs. (6)-(9) are intended as recursive least-squares updates of the same Gram inverse, so the claimed equivalence is a mathematical identity, not a quantity fitted to its own prediction target. The AFAM stores sums of outer products S'^T S' and never stores labels or predictions, so no fitted input is renamed as a prediction. The central identity is inherited from the externally published ACIL result (Zhuang et al., 2022), and although one author overlaps, that citation is not the only support offered and is independently falsifiable. The acknowledged limitation that the frozen extractor 'might not be optimal for every domain' is an assumption about transfer, not circularity. The main verifiable defect is a correctness problem rather than a circular one: Eq. (7) as printed is missing an operator between the second and third terms, and Appendix Eq. (b) writes A_t[Q_{t-1} - S_t'^T Y_t] where Eq. (5) requires a plus sign, so the appendix does not actually establish the asserted identity. That is an unsupported or misstated proof, but it is not the target result being used to derive itself.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The analysis rests on four free hyperparameters, including the expansion size that is tuned on the reported benchmarks, and on four mathematical or domain assumptions. The most fragile assumption is that frozen task-0 speech features generalize to all future keywords. No new physical or ontological entities are introduced; the AFAM is a correlation matrix and the AFE is a random projection, both mathematical objects. The paper does not ship code, data, or machine-checked certificates, so all verification of the central equivalence is the reader's reconstruction.

free parameters (4)
  • AFE expansion size = 128, 256, and 512 in experiments; main results use 128/256, best ablation uses 512
    The acoustic feature expansion dimension is a hand-chosen hyperparameter. Table 3 shows accuracy rises monotonically with expansion size (86.57 without AFE to 89.68 with 512), so the reported headline numbers depend on this choice.
  • Regularization coefficient gamma = not reported
    Gamma appears in Eqs. (2)-(6) as the ridge penalty in the least-squares solution. Its value is never stated, yet it controls invertibility and generalization of the analytic classifier.
  • AFE random initialization seed = not reported
    The random projection weights in Sec. 3.2.2 are drawn from a normal distribution and kept fixed, but the distribution parameters and seed are not specified, so the exact expanded features are not reproducible.
  • Task-0 pretraining epochs and optimizer hyperparameters = not explicitly stated; the paper says 'multiple epochs' and later uses 50 epochs for tasks
    The quality of the frozen feature extractor depends on the task-0 training setup. Without exact values, a replication cannot guarantee the same feature space.
assumptions (4)
  • standard math Woodbury matrix identity
    Invoked in Appendix A to derive the recursive update for the AFAM matrix. This is unproved background linear algebra and is standard.
  • domain assumption Task-0 feature extractor transfers to unseen keywords
    Sec. 3.2.1 freezes the CNN backbone after pretraining on task 0. All future keywords must be linearly separable in that fixed feature space, otherwise no later stage can compensate.
  • ad hoc to paper Random AFE expansion preserves or enhances linear separability
    Sec. 3.2.2 inserts a random linear expansion and Table 3 shows it is needed for top accuracy. There is no derivation of why a fixed random projection improves separability; it is empirically assumed.
  • domain assumption Squared-error least-squares objective is an adequate surrogate for the cross-entropy objective in Eq. (1)
    The problem is formulated with cross-entropy, but the analytic solution optimizes a ridge-regression objective. The paper does not prove these objectives have the same optimum; the empirical results are the only support.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AnalyticKWS: Towards Exemplar-Free Analytic Class Incremental Learning for Small-footprint Keyword Spotting." pith.science (2026). https://pith.science/paper/RXTWKIVJ

@misc{pith2026250511817,
  author       = {Pith},
  title        = {Pith review of: AnalyticKWS: Towards Exemplar-Free Analytic Class Incremental Learning for Small-footprint Keyword Spotting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RXTWKIVJ}},
  note         = {Machine review of arXiv:2505.11817}
}
read the original abstract

Keyword spotting (KWS) offers a vital mechanism to identify spoken commands in voice-enabled systems, where user demands often shift, requiring models to learn new keywords continually over time. However, a major problem is catastrophic forgetting, where models lose their ability to recognize earlier keywords. Although several continual learning methods have proven their usefulness for reducing forgetting, most existing approaches depend on storing and revisiting old data to combat catastrophic forgetting. Though effective, these methods face two practical challenges: 1) privacy risks from keeping user data and 2) large memory and time consumption that limit deployment on small devices. To address these issues, we propose an exemplar-free Analytic Continual Learning (AnalyticKWS) method that updates model parameters without revisiting earlier data. Inspired by efficient learning principles, AnalyticKWS computes a closed-form analytical solution for model updates and requires only a single epoch of adaptation for incoming keywords. AnalyticKWS demands fewer computational resources by avoiding gradient-based updates and does not store old data. By eliminating the need for back-propagation during incremental learning, the model remains lightweight and efficient. As a result, AnalyticKWS meets the challenges mentioned earlier and suits resource-limited settings well. Extensive experiments on various datasets and settings show that AnalyticKWS consistently outperforms existing continual learning methods.

Figures

Figures reproduced from arXiv: 2505.11817 by the authors.

Figure 1
Figure 1. An overview of the AnalyticKWS method: (a) Train the whole model on the first task for multiple epochs to get a strong feature extractor, then (b) Apply analytic re-alignment for one epoch to increase the pre￾classifier feature dimension. Next, proceed to the incremental keywords stage, where the model trains for one epoch per new task, assisted by a correlation matrix AFAM (Eq. (6)) that encodes past knowledge. Thi… view at source ↗
Figure 2
Figure 2. Task-wise performance comparison of different methods with 500 buffer size. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Task-wise accuracy on GSC-v2 with 11 tasks. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Task-wise accuracy on GSC-v2 with 21 tasks. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 36 canonical work pages

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Abhijeet Awasthi, Kevin Kilgour, and Hassan Rom. 2021. Teaching keyword spotters to spot new keywords with limited examples. In Proc. Interspeech

    work page 2021

  4. [4]

    Juan Pablo Bello, Charlie Mydlarz, and Justin Salamon. 2018. Sound Analysis in Smart Cities . Springer International Publishing, pages 373--397

    work page 2018

  5. [5]

    Eden Belouadah and Adrian Popescu. 2019. Il2m: Class incremental learning with dual memory. In Proc. the IEEE/CVF International Conference on Computer Vision, pages 583--592

    work page 2019

  6. [6]

    Eden Belouadah, Adrian Popescu, and Ioannis Kanellos. 2021. A Comprehensive Study of Class Incremental Learning Algorithms for Visual Tasks . Neural Networks, 135:38--54

    work page 2021

  7. [7]

    Umberto Cappellazzo, Muqiao Yang, Daniele Falavigna, and Alessio Brutti. 2023. Sequence-level knowledge distillation for class-incremental end-to-end spoken language understanding. Proceedings of Interspeech

    work page 2023

  8. [8]

    Arslan Chaudhry, Puneet K Dokania, Thalaiyasingam Ajanthan, and Philip HS Torr. 2018. Riemannian walk for incremental learning: Understanding forgetting and intransigence. In Proceedings of the European conference on computer vision (ECCV), pages 532--547

    work page 2018

Show all 45 references
  1. [9]

    Chen Chen, Ruizhe Li, Yuchen Hu, Yuanyuan Chen, Chengwei Qin, and Qiang Zhang. 2024. Overcoming catastrophic forgetting by exemplar selection in task-oriented dialogue system. arXiv preprint arXiv:2405.10992

    2024 arXiv

  2. [10]

    Guoguo Chen, Carolina Parada, and Georg Heigold. 2014. Small-footprint keyword spotting using deep neural networks. In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 4087--4091

    2014

  3. [11]

    Seungwoo Choi, Seokjun Seo, Beomjun Shin, Hyeongmin Byun, Martin Kersner, Beomsu Kim, Dongyoung Kim, and Sungjoo Ha. 2019. Temporal convolution for real-time keyword spotting on mobile devices. In Proc. Interspeech

    2019

  4. [12]

    Gene H Golub and Charles F Van Loan. 2013. Matrix computations. JHU press

    2013

  5. [13]

    Dipam Goswami, Yuyang Liu, Bart Twardowski, and Joost van de Weijer. 2024 a . Fecam: Exploiting the heterogeneity of class distributions in exemplar-free continual learning. Advances in Neural Information Processing Systems, 36

    2024

  6. [14]

    Dipam Goswami, Yuyang Liu, Bart omiej Twardowski, and Joost van de Weijer. 2024 b . Fecam: Exploiting the heterogeneity of class distributions in exemplar-free continual learning. Advances in Neural Information Processing Systems, 36

    2024

  7. [15]

    Dipam Goswami, Albin Soutif-Cormerais, Yuyang Liu, Sandesh Kamath, Bart Twardowski, Joost van de Weijer, et al. 2024 c . Resurrecting old classes with new data for exemplar-free continual learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

    2024

  8. [16]

    Dipam Goswami, Albin Soutif-Cormerais, Yuyang Liu, Sandesh Kamath, Bart Twardowski, Joost van de Weijer, et al. 2024 d . Resurrecting old classes with new data for exemplar-free continual learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogn...

    2024

  9. [17]

    Saihui Hou, Xinyu Pan, Chen Change Loy, Zilei Wang, and Dahua Lin. 2019. Learning a Unified Classifier Incrementally via Rebalancing . In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 831--839

    2019

  10. [18]

    Yizheng Huang, Nana Hou, and Nancy F Chen. 2022. Progressive continual learning for spoken keyword spotting. In Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7552--7556

    2022

  11. [19]

    Byeonggeun Kim, Simyung Chang, Jinkyu Lee, and Dooyong Sung. 2021. Broadcasted residual learning for efficient keyword spotting. In Proc. Interspeech

    2021

  12. [20]

    James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. 2017. Overcoming Catastrophic Forgetting in Neural Networks . Proceedings of the National Academy o...

    2017

  13. [21]

    Iv \'a n L \'o pez-Espejo, Zheng-Hua Tan, John HL Hansen, and Jesper Jensen. 2021. Deep spoken keyword spotting: An overview. IEEE Access, 10:4169--4199

    2021

  14. [22]

    David Lopez-Paz and Marc'Aurelio Ranzato. 2017. Gradient episodic memory for continual learning. Advances in neural information processing systems, 30

    2017

  15. [23]

    Anupam Mandal, KR Prasanna Kumar, and Pabitra Mitra. 2014. Recent developments in spoken term detection: a survey. International Journal of Speech Technology, 17:183--198

    2014

  16. [24]

    Mark Mazumder, Colby Banbury, Josh Meyer, Pete Warden, and Vijay Janapa Reddi. 2021. Few-shot keyword spotting in any language. In Proc. Interspeech

    2021

  17. [25]

    Michael McCloskey and Neal J Cohen. 1989. Catastrophic interference in connectionist networks: The sequential learning problem. In Psychology of learning and motivation, volume 24, pages 109--165. Elsevier

    1989

  18. [26]

    Dianwen Ng, Yang Xiao, Jia Qi Yip, Zhao Yang, Biao Tian, Qiang Fu, Eng Siong Chng, and Bin Ma. 2023. Small footprint multi-channel network for keyword spotting with centroid based awareness. In Proc. Interspeech, pages 296--300

    2023

  19. [27]

    German I Parisi, Ronald Kemker, Jose L Part, Christopher Kanan, and Stefan Wermter. 2019. Continual Lifelong Learning with Neural Networks: A Review . Neural networks, 113:54--71

    2019

  20. [28]

    Archit Parnami and Minwoo Lee. 2022. Few-shot keyword spotting with prototypical networks. In Proc. International Conference on Machine Learning Technologies (ICMLT), pages 277--283

    2022

  21. [29]

    Francesco Pelosin, Saurav Jha, Andrea Torsello, Bogdan Raducanu, and Joost van de Weijer. 2022. Towards exemplar-free continual learning in vision transformers: an account of attention, functional and weight regularization. In Proceedings of the IEEE/CVF Conference on Computer...

    2022

  22. [30]

    Tianyi Peng and Yang Xiao. 2024. Dark experience for incremental keyword spotting. arXiv preprint:2409.08153

    2024 arXiv

  23. [31]

    Gr \'e goire Petit, Adrian Popescu, Hugo Schindler, David Picard, and Bertrand Delezoide. 2023. Fetril: Feature translation for exemplar-free class-incremental learning. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 3911--3920

    2023

  24. [32]

    Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, Georg Sperl, and Christoph H Lampert. 2017. icarl: Incremental Classifier and Representation Learning . In Proc. the IEEE conference on Computer Vision and Pattern Recognition (CVPR), pages 2001--2010

    2017

  25. [33]

    Hippolyt Ritter, Aleksandar Botev, and David Barber. 2018. Online structured laplace approximations for overcoming catastrophic forgetting. Advances in Neural Information Processing Systems, 31

    2018

  26. [34]

    Zeyang Song, Qianhui Liu, Qu Yang, Yizhou Peng, and Haizhou Li. 2024. Ed-skws: Early-decision spiking neural networks for rapid, and energy-efficient keyword spotting. arXiv preprint arXiv:2406.12726

    2024 arXiv

  27. [35]

    Raphael Tang and Jimmy Lin. 2018. Deep residual learning for small-footprint keyword spotting. In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 5484--5488. IEEE

    2018

  28. [36]

    Yue Wu, Yinpeng Chen, Lijuan Wang, Yuancheng Ye, Zicheng Liu, Yandong Guo, and Yun Fu. 2019. Large scale incremental learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 374--382

    2019

  29. [37]

    Yang Xiao and Rohan Kumar Das. 2024 a . Configurable DOA Estimation using Incremental Learning . arXiv preprint:2407.03661

    2024 arXiv

  30. [38]

    Yang Xiao and Rohan Kumar Das. 2024 b . UCIL: An Unsupervised Class Incremental Learning Approach for Sound Event Detection . arXiv:2407.03657

    2024 arXiv

  31. [39]

    Yang Xiao, Nana Hou, and Eng Siong Chng. 2022 a . https://doi.org/10.21437/Interspeech.2022-10500 Rainbow Keywords: Efficient Incremental Learning for Online Spoken Keyword Spotting . In Proc. Interspeech, pages 3764--3768

    2022 doi

  32. [40]

    Yang Xiao, Xubo Liu, James King, Arshdeep Singh, Eng Siong Chng, Mark D Plumbley, and Wenwu Wang. 2022 b . Continual Learning for On-device Environmental Sound Classification . In Proc. the Detection and Classification of Acoustic Scenes and Events Workshop (DCASE)

    2022

  33. [41]

    Muqiao Yang, Ian Lane, and Shinji Watanabe. 2022 a . Online continual learning of end-to-end speech recognition models. Proceedings of Interspeech

    2022

  34. [42]

    Seunghan Yang, Byeonggeun Kim, Inseop Chung, and Simyung Chang. 2022 b . Personalized keyword spotting through multi-task learning. arXiv preprint arXiv:2206.13708

    2022 arXiv

  35. [43]

    Yundong Zhang, Naveen Suda, Liangzhen Lai, and Vikas Chandra. 2018. Hello edge: Keyword spotting on microcontrollers. arXiv:1711.07128

    2018 arXiv

  36. [44]

    Huiping Zhuang, Zhiping Lin, and Kar-Ann Toh. 2021. Blockwise recursive moore--penrose inverse for network learning. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 52(5):3237--3250

    2021

  37. [45]

    Huiping Zhuang, Zhenyu Weng, Hongxin Wei, Renchunzi Xie, Kar-Ann Toh, and Zhiping Lin. 2022. Acil: Analytic class-incremental learning with absolute memorization and privacy protection. Advances in Neural Information Processing Systems, 35:11602--11614

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.