The reviewed record of science sign in
Pith

arxiv: 2605.27409 · v1 · pith:F2MHOUBW · submitted 2026-05-12 · cs.NE · cs.AI· cs.LG

STARS: Spike Tail-Aware Relational Synthesis for ANN-to-SNN Data-Free Knowledge Distillation

Shuhan Ye , Yi Yu , Qixin Zhang , Hui Lu , Jiaming He , Qinggang Zhang , Li Shen , Xudong Jiang This is my paper

Reviewed by Pith2026-06-30 22:25 UTCgrok-4.3pith:F2MHOUBWopen to challenge →

classification cs.NE cs.AIcs.LG
keywords data-free knowledge distillationspiking neural networksANN-to-SNNsynthetic data generationrelational consistencytail-aware regularizationbatch normalization matching
0
0 comments X

The pith

STARS augments BN-guided data synthesis with relational alignment and tail regularization to improve ANN-to-SNN distillation without original data.

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

The paper seeks to establish that standard data-free knowledge distillation, which matches batch normalization statistics from the ANN teacher, leaves synthetic data under-constrained for SNN students whose outputs depend on threshold crossings. It introduces two additional objectives during synthesis to enforce cross-sample relational consistency and to regularize tail probabilities relevant to those thresholds. A sympathetic reader would care because many practical settings lack access to the original training images, yet SNNs are valued for their energy efficiency; closing the accuracy gap in such settings would make neuromorphic hardware more competitive. If the claim holds, synthetic batches become valid for the teacher while carrying more useful signals for the student, leading to measurable accuracy lifts on standard image benchmarks.

Core claim

STARS augments standard BN-guided synthesis with Relational Consistency Alignment, which preserves cross-sample relational consistency between teacher and student, and Tail-Aware Regularization, which regularizes threshold-relevant tail probabilities through soft exceedance over teacher-derived thresholds. Together these objectives generate synthetic batches that remain teacher-valid while becoming more informative for SNN students. Experiments on CIFAR-10, CIFAR-100, and Tiny-ImageNet across multiple ANN-SNN pairs show consistent gains over conventional DFKD baselines and even over several KD methods that use real data.

What carries the argument

Spike Tail-Aware Relational Synthesis (STARS), the synthesis procedure that augments BN matching with Relational Consistency Alignment and Tail-Aware Regularization to produce SNN-oriented surrogate batches.

If this is right

  • The method yields accuracy gains of up to 4.6 percent on CIFAR-10 and 6.7 percent on CIFAR-100 over standard DFKD baselines.
  • Performance improves consistently across multiple ANN-SNN architecture pairs and on Tiny-ImageNet as well.
  • The resulting SNNs can surpass the accuracy of several knowledge-distillation approaches that require the original training images.
  • The approach is plug-and-play, requiring only the addition of the two objectives to existing BN-guided synthesis pipelines.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • DFKD pipelines may need student-specific constraints rather than teacher-only statistics when the student uses fundamentally different activation mechanisms.
  • The same relational and tail objectives could be tested on other threshold-based or sparse student models beyond SNNs.
  • If the gains hold under stricter privacy constraints, the technique could reduce reliance on real data in regulated deployment scenarios.

Load-bearing premise

That the added Relational Consistency Alignment and Tail-Aware Regularization objectives produce synthetic batches that remain valid for the ANN teacher yet become substantially more informative for the SNN student's threshold-crossing dynamics.

What would settle it

A controlled run in which the two new objectives are added to synthesis, teacher validity is preserved, yet the resulting SNN accuracy shows no gain or a clear drop relative to plain BN-guided synthesis on the same ANN-SNN pair and dataset.

Figures

Figures reproduced from arXiv: 2605.27409 by Hui Lu, Jiaming He, Li Shen, Qinggang Zhang, Qixin Zhang, Shuhan Ye, Xudong Jiang, Yi Yu.

Figure 1
Figure 1. Figure 1: Overview of our proposed method STARS. BN-guided synthesis provides teacher-valid [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: t-SNE visualization of SNN student (S-VGG-16) penultimate-layer features under NAYER [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
read the original abstract

SNNs promise energy-efficient and low-latency inference, but their performance still trails that of ANNs. ANN-to-SNN knowledge distillation helps narrow this gap, yet the original training data are often unavailable in practical deployment settings. Existing data-free knowledge distillation (DFKD) methods synthesize surrogate data by matching teacher-side priors, especially BN statistics, but these ANN-oriented constraints mainly regularize mean and variance and therefore remain under-constrained for SNN students whose responses depend on threshold-crossing dynamics. In this paper, we propose Spike Tail-Aware Relational Synthesis (STARS), a plug-and-play method for ANN-to-SNN DFKD that augments standard BN-guided synthesis with two complementary objectives: Relational Consistency Alignment, which preserves cross-sample relational consistency between teacher and student, and Tail-Aware Regularization, which regularizes threshold-relevant tail probabilities through soft exceedance over teacher-derived thresholds. Together, these objectives generate synthetic batches that remain teacher-valid while becoming more informative for SNN students. Experiments on CIFAR-10, CIFAR-100, and Tiny-ImageNet across multiple ANN-SNN pairs show that our method consistently improves conventional DFKD baselines and even surpasses several KD methods, with gains of up to 4.6\% on CIFAR-10 and 6.7\% on CIFAR-100, highlighting the importance of complementing BN matching with relational and tail-aware constraints in SNN-oriented DFKD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes Spike Tail-Aware Relational Synthesis (STARS), a plug-and-play augmentation to data-free knowledge distillation (DFKD) from ANNs to SNNs. Standard BN-statistics matching during surrogate data synthesis is extended with Relational Consistency Alignment (to preserve cross-sample relational consistency between teacher and student) and Tail-Aware Regularization (to regularize threshold-relevant tail probabilities via soft exceedance over teacher-derived thresholds). The central claim is that the resulting synthetic batches remain valid for the ANN teacher yet become substantially more informative for the SNN student's threshold-crossing dynamics. Experiments across CIFAR-10, CIFAR-100 and Tiny-ImageNet with multiple ANN-SNN pairs are reported to yield consistent gains over conventional DFKD baselines (up to 4.6 % on CIFAR-10 and 6.7 % on CIFAR-100) and to surpass several KD methods.

Significance. If the teacher-validity of the generated data is demonstrated and the reported gains are reproducible, the work would provide a practical route to adapt DFKD pipelines to SNNs by explicitly addressing the mismatch between ANN mean/variance priors and SNN spike-threshold dynamics. The plug-and-play nature and the empirical outperformance of some supervised KD baselines would be of direct interest to the neuromorphic and efficient-inference communities.

major comments (2)
  1. [Abstract] Abstract: The claim that the augmented objectives produce synthetic batches that 'remain teacher-valid' is unsupported by any reported measurement of post-optimization BN matching error, teacher accuracy on the generated data, or analysis of the relative weighting that keeps the BN constraint intact. Because the synthesis objective is a weighted sum that now includes the new Relational Consistency Alignment and Tail-Aware Regularization terms, this verification is load-bearing for the central claim that the batches stay valid for the ANN teacher.
  2. [Experiments] Experiments (results paragraph): The stated gains of up to 4.6 % on CIFAR-10 and 6.7 % on CIFAR-100 are presented without reference to experimental protocol, baseline definitions, error bars, statistical tests, or ablation studies isolating the contribution of each new objective. These omissions prevent assessment of whether the observed improvements are attributable to the proposed SNN-oriented constraints rather than to implementation details or hyper-parameter choices.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback highlighting the need for stronger empirical support of our claims. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the augmented objectives produce synthetic batches that 'remain teacher-valid' is unsupported by any reported measurement of post-optimization BN matching error, teacher accuracy on the generated data, or analysis of the relative weighting that keeps the BN constraint intact. Because the synthesis objective is a weighted sum that now includes the new Relational Consistency Alignment and Tail-Aware Regularization terms, this verification is load-bearing for the central claim that the batches stay valid for the ANN teacher.

    Authors: We agree that explicit verification is required to substantiate the teacher-validity claim. The revised manuscript will add quantitative results on post-optimization BN matching error, teacher accuracy evaluated on the generated batches, and an analysis of the relative weighting between the BN term and the new objectives to confirm that the BN constraint remains dominant and the synthetic data stays valid for the ANN teacher. revision: yes

  2. Referee: [Experiments] Experiments (results paragraph): The stated gains of up to 4.6 % on CIFAR-10 and 6.7 % on CIFAR-100 are presented without reference to experimental protocol, baseline definitions, error bars, statistical tests, or ablation studies isolating the contribution of each new objective. These omissions prevent assessment of whether the observed improvements are attributable to the proposed SNN-oriented constraints rather than to implementation details or hyper-parameter choices.

    Authors: We acknowledge these reporting gaps. The revised manuscript will expand the experimental section to include the full protocol details, explicit baseline definitions, error bars from multiple independent runs, statistical significance tests, and dedicated ablation studies that isolate the individual contributions of Relational Consistency Alignment and Tail-Aware Regularization. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical method augmentation with independent experimental validation

full rationale

The paper introduces STARS as an augmentation of standard BN-guided data synthesis with two new objectives (Relational Consistency Alignment and Tail-Aware Regularization). No equations, derivations, or first-principles claims appear that reduce the reported accuracy gains to quantities defined by the method itself or to self-citation chains. The central results are empirical improvements measured on CIFAR-10/100 and Tiny-ImageNet across ANN-SNN pairs; these outcomes are not forced by construction from the synthesis objectives. The method description remains self-contained against external benchmarks and does not invoke uniqueness theorems or fitted parameters renamed as predictions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities are stated or can be inferred from the provided text.

pith-pipeline@v0.9.1-grok · 5818 in / 1067 out tokens · 29301 ms · 2026-06-30T22:25:48.240285+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages · 4 internal anchors

  1. [1]

    Networks of spiking neurons: the third generation of neural network models

    Wolfgang Maass. Networks of spiking neurons: the third generation of neural network models. Neural networks, 10(9):1659–1671, 1997

    work page 1997

  2. [2]

    Surrogate gradient learning in spiking neural networks.IEEE Signal Processing Magazine, 36(6):51–63, 2019

    Emre O Neftci, Hesham Mostafa, and Friedemann Zenke. Surrogate gradient learning in spiking neural networks.IEEE Signal Processing Magazine, 36(6):51–63, 2019

    work page 2019

  3. [3]

    Towards spike-based machine intelligence with neuromorphic computing.Nature, 575(7784):607–617, 2019

    Kaushik Roy, Akhilesh Jaiswal, and Priyadarshini Panda. Towards spike-based machine intelligence with neuromorphic computing.Nature, 575(7784):607–617, 2019

    work page 2019

  4. [4]

    Spatio-temporal backpropagation for training high-performance spiking neural networks.Frontiers in neuroscience, 12:331, 2018

    Yujie Wu, Lei Deng, Guoqi Li, Jun Zhu, and Luping Shi. Spatio-temporal backpropagation for training high-performance spiking neural networks.Frontiers in neuroscience, 12:331, 2018

    work page 2018

  5. [5]

    Going deeper with directly-trained larger spiking neural networks

    Hanle Zheng, Yujie Wu, Lei Deng, Yifan Hu, and Guoqi Li. Going deeper with directly-trained larger spiking neural networks. InProceedings of the AAAI conference on artificial intelligence, volume 35, pages 11062–11070, 2021

    work page 2021

  6. [6]

    Loihi: A neuromorphic manycore processor with on-chip learning.Ieee micro, 38(1):82–99, 2018

    Mike Davies, Narayan Srinivasa, Tsung-Han Lin, Gautham Chinya, Yongqiang Cao, Sri Harsha Choday, Georgios Dimou, Prasad Joshi, Nabil Imam, Shweta Jain, et al. Loihi: A neuromorphic manycore processor with on-chip learning.Ieee micro, 38(1):82–99, 2018

    work page 2018

  7. [7]

    Incorporating learnable membrane time constant to enhance learning of spiking neural networks

    Wei Fang, Zhaofei Yu, Yanqi Chen, Timothée Masquelier, Tiejun Huang, and Yonghong Tian. Incorporating learnable membrane time constant to enhance learning of spiking neural networks. InProceedings of the IEEE/CVF international conference on computer vision, pages 2661–2671, 2021

    work page 2021

  8. [8]

    Optimal conversion of conventional artificial neural networks to spiking neural networks.arXiv preprint arXiv:2103.00476, 2021

    Shikuang Deng and Shi Gu. Optimal conversion of conventional artificial neural networks to spiking neural networks.arXiv preprint arXiv:2103.00476, 2021

    work page arXiv 2021

  9. [9]

    Ucf-crime- dvs: A novel event-based dataset for video anomaly detection with spiking neural networks

    Yuanbin Qian, Shuhan Ye, Chong Wang, Xiaojie Cai, Jiangbo Qian, and Jiafei Wu. Ucf-crime- dvs: A novel event-based dataset for video anomaly detection with spiking neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 6577–6585, 2025

    work page 2025

  10. [10]

    Differ- entiable spike: Rethinking gradient-descent for training spiking neural networks.Advances in neural information processing systems, 34:23426–23439, 2021

    Yuhang Li, Yufei Guo, Shanghang Zhang, Shikuang Deng, Yongqing Hai, and Shi Gu. Differ- entiable spike: Rethinking gradient-descent for training spiking neural networks.Advances in neural information processing systems, 34:23426–23439, 2021

    work page 2021

  11. [11]

    Glif: A unified gated leaky integrate-and- fire neuron for spiking neural networks.Advances in Neural Information Processing Systems, 35:32160–32171, 2022

    Xingting Yao, Fanrong Li, Zitao Mo, and Jian Cheng. Glif: A unified gated leaky integrate-and- fire neuron for spiking neural networks.Advances in Neural Information Processing Systems, 35:32160–32171, 2022

    work page 2022

  12. [12]

    Spikformer: When spiking neural network meets transformer

    Zhaokun Zhou, Yuesheng Zhu, Chao He, Yaowei Wang, Shuicheng YAN, Yonghong Tian, and Li Yuan. Spikformer: When spiking neural network meets transformer. InThe Eleventh International Conference on Learning Representations

    work page

  13. [13]

    Progressive tandem learning for pattern recognition with deep spiking neural networks.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(11):7824–7840, 2021

    Jibin Wu, Chenglin Xu, Xiao Han, Daquan Zhou, Malu Zhang, Haizhou Li, and Kay Chen Tan. Progressive tandem learning for pattern recognition with deep spiking neural networks.IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(11):7824–7840, 2021

    work page 2021

  14. [14]

    Training spiking neural networks with local tandem learning.Advances in Neural Information Processing Systems, 35:12662–12676, 2022

    Qu Yang, Jibin Wu, Malu Zhang, Yansong Chua, Xinchao Wang, and Haizhou Li. Training spiking neural networks with local tandem learning.Advances in Neural Information Processing Systems, 35:12662–12676, 2022

    work page 2022

  15. [15]

    Temporal- coded spiking neural networks with dynamic firing threshold: Learning with event-driven backpropagation

    Wenjie Wei, Malu Zhang, Hong Qu, Ammar Belatreche, Jian Zhang, and Hong Chen. Temporal- coded spiking neural networks with dynamic firing threshold: Learning with event-driven backpropagation. InProceedings of the IEEE/CVF international conference on computer vision, pages 10552–10562, 2023

    work page 2023

  16. [16]

    Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural net- works

    Yufei Guo, Xinyi Tong, Yuanpei Chen, Liwen Zhang, Xiaode Liu, Zhe Ma, and Xuhui Huang. Recdis-snn: Rectifying membrane potential distribution for directly training spiking neural net- works. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 326–335, 2022. 10

    work page 2022

  17. [17]

    Distilling the Knowledge in a Neural Network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  18. [18]

    FitNets: Hints for Thin Deep Nets

    Adriana Romero, Nicolas Ballas, Samira Ebrahimi Kahou, Antoine Chassang, Carlo Gatta, and Yoshua Bengio. Fitnets: hints for thin deep nets (2014).arXiv preprint arXiv:1412.6550, 3, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  19. [19]

    Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer

    Sergey Zagoruyko and Nikos Komodakis. Paying more attention to attention: Improving the performance of convolutional neural networks via attention transfer.arXiv preprint arXiv:1612.03928, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  20. [20]

    Relational knowledge distillation

    Wonpyo Park, Dongju Kim, Yan Lu, and Minsu Cho. Relational knowledge distillation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3967–3976, 2019

    work page 2019

  21. [21]

    Contrastive Representation Distillation

    Yonglong Tian, Dilip Krishnan, and Phillip Isola. Contrastive representation distillation.arXiv preprint arXiv:1910.10699, 2019

    work page Pith review arXiv 1910

  22. [22]

    Joint A-SNN: Joint training of artificial and spiking neural networks via self-distillation and weight factorization.Pattern Recognition, 142:109639, 2023

    Yufei Guo, Weihang Peng, Yuanpei Chen, Liwen Zhang, Xiaode Liu, Xuhui Huang, and Zhe Ma. Joint A-SNN: Joint training of artificial and spiking neural networks via self-distillation and weight factorization.Pattern Recognition, 142:109639, 2023

    work page 2023

  23. [23]

    BKDSNN: Enhancing the performance of learning-based spiking neural networks training with blurred knowledge distillation

    Zekai Xu, Kang You, Qinghai Guo, Xiang Wang, and Zhezhi He. BKDSNN: Enhancing the performance of learning-based spiking neural networks training with blurred knowledge distillation. InEuropean Conference on Computer Vision, pages 106–123. Springer, 2024

    work page 2024

  24. [24]

    Cross knowledge distillation between artificial and spiking neural networks

    Shuhan Ye, Yuanbin Qian, Chong Wang, Sunqi Lin, Jiazhen Xu, Jiangbo Qian, and Yuqi Li. Cross knowledge distillation between artificial and spiking neural networks. In2025 IEEE International Conference on Multimedia and Expo (ICME), pages 1–6. IEEE, 2025

    work page 2025

  25. [25]

    Temporal separation with entropy regularization for knowledge distillation in spiking neural networks

    Kairong Yu, Chengting Yu, Tianqing Zhang, Xiaochen Zhao, Shu Yang, Hongwei Wang, Qiang Zhang, and Qi Xu. Temporal separation with entropy regularization for knowledge distillation in spiking neural networks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8806–8816, 2025

    work page 2025

  26. [26]

    Efficient ANN-guided distillation: Aligning rate-based features of spiking neural networks through hybrid block-wise replacement

    Shu Yang, Chengting Yu, Lei Liu, Hanzhi Ma, Aili Wang, and Erping Li. Efficient ANN-guided distillation: Aligning rate-based features of spiking neural networks through hybrid block-wise replacement. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10025–10035, 2025

    work page 2025

  27. [27]

    Knowledge extraction with no observable data.Advances in Neural Information Processing Systems, 32, 2019

    Jaemin Yoo, Minyong Cho, Taebum Kim, and U Kang. Knowledge extraction with no observable data.Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  28. [28]

    Data-free adversarial distillation.arXiv preprint arXiv:1912.11006, 2019

    Gongfan Fang, Jie Song, Chengchao Shen, Xinchao Wang, Da Chen, and Mingli Song. Data-free adversarial distillation.arXiv preprint arXiv:1912.11006, 2019

    work page arXiv 1912

  29. [29]

    Robust and resource-efficient data-free knowledge distillation by generative pseudo replay

    Kuluhan Binici, Shivam Aggarwal, Nam Trung Pham, Karianto Leman, and Tulika Mitra. Robust and resource-efficient data-free knowledge distillation by generative pseudo replay. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 6089–6096, 2022

    work page 2022

  30. [30]

    Zero-shot knowledge transfer via adversarial belief matching

    Paul Micaelli and Amos J Storkey. Zero-shot knowledge transfer via adversarial belief matching. Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  31. [31]

    Zero-shot knowledge distillation in deep networks

    Gaurav Kumar Nayak, Konda Reddy Mopuri, Vaisakh Shaj, Venkatesh Babu Radhakrishnan, and Anirban Chakraborty. Zero-shot knowledge distillation in deep networks. InInternational conference on machine learning, pages 4743–4751. PMLR, 2019

    work page 2019

  32. [32]

    Data-free learning of student networks

    Hanting Chen, Yunhe Wang, Chang Xu, Zhaohui Yang, Chuanjian Liu, Boxin Shi, Chunjing Xu, Chao Xu, and Qi Tian. Data-free learning of student networks. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 3514–3522, 2019. 11

    work page 2019

  33. [33]

    Dreaming to distill: Data-free knowledge transfer via deepinversion

    Hongxu Yin, Pavlo Molchanov, Jose M Alvarez, Zhizhong Li, Arun Mallya, Derek Hoiem, Niraj K Jha, and Jan Kautz. Dreaming to distill: Data-free knowledge transfer via deepinversion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8715–8724, 2020

    work page 2020

  34. [34]

    Contrastive model inversion for data-free knowledge distillation

    Gongfan Fang, Jie Song, Xinchao Wang, Chengchao Shen, Xingen Wang, and Mingli Song. Contrastive model inversion for data-free knowledge distillation. 2021

    work page 2021

  35. [35]

    Nayer: Noisy layer data generation for efficient and effective data-free knowledge distillation

    Minh-Tuan Tran, Trung Le, Xuan-May Le, Mehrtash Harandi, Quan Hung Tran, and Dinh Phung. Nayer: Noisy layer data generation for efficient and effective data-free knowledge distillation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 23860–23869, 2024

    work page 2024

  36. [36]

    Architecture, dataset and model-scale agnostic data-free meta-learning

    Zixuan Hu, Li Shen, Zhenyi Wang, Tongliang Liu, Chun Yuan, and Dacheng Tao. Architecture, dataset and model-scale agnostic data-free meta-learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7736–7745, 2023

    work page 2023

  37. [37]

    Learning to learn from apis: Black-box data-free meta-learning

    Zixuan Hu, Li Shen, Zhenyi Wang, Baoyuan Wu, Chun Yuan, and Dacheng Tao. Learning to learn from apis: Black-box data-free meta-learning. InInternational Conference on Machine Learning, pages 13610–13627. PMLR, 2023

    work page 2023

  38. [38]

    Free: Faster and better data-free meta-learning

    Yongxian Wei, Zixuan Hu, Zhenyi Wang, Li Shen, Chun Yuan, and Dacheng Tao. Free: Faster and better data-free meta-learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 23273–23282, 2024

    work page 2024

  39. [39]

    Task groupings regularization: Data-free meta-learning with heterogeneous pre-trained models

    Yongxian Wei, Zixuan Hu, Li Shen, Zhenyi Wang, Yu Li, Chun Yuan, and Dacheng Tao. Task groupings regularization: Data-free meta-learning with heterogeneous pre-trained models. 2024

    work page 2024

  40. [40]

    Task-distributionally robust data-free meta-learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    Zixuan Hu, Yongxian Wei, Li Shen, Zhenyi Wang, Baoyuan Wu, Chun Yuan, and Dacheng Tao. Task-distributionally robust data-free meta-learning.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

  41. [41]

    Cambridge University Press, 2014

    Wulfram Gerstner, Werner M Kistler, Richard Naud, and Liam Paninski.Neuronal dynamics: From single neurons to networks and models of cognition. Cambridge University Press, 2014

    work page 2014

  42. [42]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layers of features from tiny images. 2009

    work page 2009

  43. [43]

    Very Deep Convolutional Networks for Large-Scale Image Recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition.arXiv preprint arXiv:1409.1556, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  44. [44]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    work page 2016

  45. [45]

    Tiny imagenet visual recognition challenge.CS 231N, 7(7):3, 2015

    Yann Le, Xuan Yang, et al. Tiny imagenet visual recognition challenge.CS 231N, 7(7):3, 2015

    work page 2015

  46. [46]

    Up to 100x faster data-free knowledge distillation

    Gongfan Fang, Kanya Mo, Xinchao Wang, Jie Song, Shitao Bei, Haofei Zhang, and Mingli Song. Up to 100x faster data-free knowledge distillation. InProceedings of the AAAI conference on artificial intelligence, volume 36, pages 6597–6604, 2022

    work page 2022

  47. [47]

    Lasnn: Layer-wise ann-to-snn distillation for effective and efficient training in deep spiking neural networks.Neurocomputing, page 131351, 2025

    Di Hong, Yu Qi, and Yueming Wang. Lasnn: Layer-wise ann-to-snn distillation for effective and efficient training in deep spiking neural networks.Neurocomputing, page 131351, 2025

    work page 2025

  48. [48]

    Knowledge distillation for spiking neural network

    Fengzhao Zhang, Chengting Yu, Hanzhi Ma, Zheming Gu, and Er-ping Li. Knowledge distillation for spiking neural network. In2023 5th International Conference on Robotics, Intelligent Control and Artificial Intelligence (RICAI), pages 1015–1020. IEEE, 2023

    work page 2023

  49. [49]

    Visualizing data using t-sne.Journal of machine learning research, 9(11), 2008

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of machine learning research, 9(11), 2008. 12 A Detailed Experimental Settings All experiments are conducted on NVIDIA RTX Pro 6000 GPUs. This appendix summarizes the detailed experimental settings used in our DFKD experiments. Unless otherwise specified, STARS is added on to...

    work page 2008

  50. [50]

    Institutional review board (IRB) approvals or equivalent for research with human subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals (or an equivalent approval/review based on the requirements of your country or ...

    work page