Download pdf
Review of Tiny Machine Learning Model Pruning Techniques on Resource-constrained Devices
- 1 Zhejiang University - University of Illinois at Urbana-Champaign Institute, Jia Xing, China
* Author to whom correspondence should be addressed.
Abstract
Lightweight pruning facilitates the deployment of machine learning models on resource-constrained devices. This review systematically examines pruning techniques across different technical paths, along with lightweight strategies that incorporate pruning. Regarding pruning techniques, the review successively delves into the principles, implementation methods, and applicable scenarios of structured pruning, unstructured pruning, and automated pruning. Structured pruning holds significant advantages in hardware implementation; unstructured pruning, on the other hand, demonstrates unique potential in fine-grained optimization, while automatic pruning methods achieve more precise model compression through intelligent search strategies. Subsequently, it centers on the synergy among pruning and other lightweight approaches, presenting the integration of pruning with quantization, the combination of pruning and distillation, as well as the pruning concepts incorporated in lightweight neural network architectures. The review concludes by highlighting some current challenges facing pruning technologies and offering insights into potential future research directions. These integration strategies not only enhance the model's operational efficiency in resource-constrained environments, but also offer innovative ideas for further compression while maintaining high accuracy.
Keywords
TinyML, pruning, quantization, distillation, lightweight neural network architecture
[1]. Dutta L & Bharali S (2021). TinyML meets IoT: A comprehensive survey. Internet of Things, 16, 100461.
[2]. Ray, P. P. (2022). A review on TinyML: State-of-the-art and prospects. Journal of King Saud University-Computer and Information Sciences, 34(4), 1595–1623.
[3]. Jiang, X., Li, Z., & Huang, L., et al. (2022). Review of neural network pruning techniques. Journal of Applied Science, 40(5), 838–849. https://doi.org/10.3969/j.issn.0255-8297.2022.05.013.
[4]. Han, S., Pool, J., Tran, J., et al. (2015, December 7–12). Learning both weights and connections for efficient neural networks. In Advances in Neural Information Processing Systems 28 (Vol. 2, pp. 1135–1143). Montreal, Canada: Neural Information Processing Systems.
[5]. Li, H., Kadav, A., Durdanovic, I., Samet, H., & Graf, H. P. (2016). Pruning filters for efficient convnets. arXiv preprint arXiv:1608.08710.
[6]. Wu, W., Fan, Q., Zurada, J. M., et al. (2014). Batch gradient method with smoothing L1/2 regularization for training of feedforward neural networks. Neural Networks: The Official Journal of the International Neural Network Society, 5072–5078. https://doi.org/10.1016/j.neunet.2013.11.006
[7]. Luo, X., Chang, X., & Ban, X. (2016). Regression and classification using extreme learning machine based on L1-norm and L2-norm. Neurocomputing, 174, 179–186.
[8]. Cheng, D., Zheng, H., & Chen, J.(2024). Deep convolutional neural network pruning method based on similarity perception. Small Microcomputer System, 45(11), 2656–2662. https://doi.org/10.20009/j.cnki.21-1106/TP.2023-0302
[9]. Wang, B. (2023). Convolutional neural network structured pruning with enhanced linear representation redundancy [Doctoral dissertation, Tianjin Normal University].
[10]. Wang, J., Ji, F., & Yuan, X. (2022). Structured pruning algorithm based on gradient tracking. Computer simulation, 39(8), 347–355, 414. https://doi.org/10.3969/j.issn.1006-9348.2022.08.067
[11]. Liu, N., Ma, X., Xu, Z., et al. (2020, February 7–12). AutoCompress: An automatic DNN structured pruning framework for ultra‐high compression rates. In Proceedings of the 34th AAAI Conference on Artificial Intelligence (pp. 4876–4883). Association for the Advancement of Artificial Intelligence.
[12]. Lin, M., Ji, R., Zhang, Y., et al. (2021, January 7–15). Channel pruning via automatic structure search. In Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence (Vol. 2 of 8, pp. 669–675). Curran Associates, Inc.
[13]. Xiao, X., Wang, Z., & Rajasekaran, S. (2020, December 8–14). AutoPrune: Automatic network pruning by regularizing auxiliary parameters. In Advances in Neural Information Processing Systems 32 (Vol. 18 of 20, 32nd Conference on Neural Information Processing Systems [NeurIPS 2019], pp. 13636–13646). Curran Associates, Inc.
[14]. Liu, Z., Mu, H., Zhang, X., et al. (2019, October 27–November 2). MetaPruning: Meta learning for automatic neural network channel pruning. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV) (pp. 3295–3304). Institute of Electrical and Electronics Engineers.
[15]. Wang, X., Liu, P., Xiang, S., et al. (2023). Unstructured pruning method based on neural architecture search. Pattern Recognition and Artificial Intelligence, 36(5), 448–458. https://doi.org/10.16451/j.cnki.issn1003-6059.202305005.
[16]. Zhang, M., Lu, Q., Li, W., et al. (2021). Deep neural network compression algorithm based on joint dynamic pruning. Computer Application, 41(6), 1589–1596. https://doi.org/10.11772/j.issn.1001-9081.2020121914
[17]. Dan, Y. (2023). Research and application of neural network quantitative perception training method [Doctoral dissertation, University of Electronic Science and Technology of China].
[18]. Ghamari, S., Ozcan, K., Dinh, T., et al. (2021). Quantization-guided training for compact TinyML models [Preprint]. arXiv. https://arxiv.org/abs/2103.06231.
[19]. Hinton, G. (2015). Distilling the knowledge in a neural network [Preprint]. arXiv. https://arxiv.org/abs/1503.02531.
[20]. Suwannaphong, T., Jovan, F., Craddock, I., et al. (2024). Optimising TinyML with quantization and distillation of transformer and mamba models for indoor localisation on edge devices [Preprint]. arXiv. https://arxiv.org/abs/2412.09289
[21]. Deng, X., Wang, Y., Luo, J., et al. (2024). Target compression algorithm based on feature enhanced knowledge distillation. Sensors and Microsystems, 43(5), 116–120. https://doi.org/10.13873/J.1000-9787(2024)05-0116-05.
[22]. Park, W., Kim, D., Lu, Y., et al. (2019). Relational knowledge distillation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 3967–3976). IEEE.
[23]. Howard, A. G. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications [Preprint]. arXiv. https://arxiv.org/abs/1704.04861
[24]. Zhang, X., Zhou, X., Lin, M., et al. (2018). Shufflenet: An extremely efficient convolutional neural network for mobile devices. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 6848–6856). IEEE.
[25]. Chen, F., Li, S., Han, J., Ren, F., & Yang, Z. (2024). Review of lightweight deep convolutional neural networks. Archives of Computational Methods in Engineering, 31(4), 1915–1937.
[26]. Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018). MobileNetV2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 4510–4520).
[27]. Tan, M., & Le, Q. (2019). EfficientNet: Rethinking model scaling for convolutional neural networks. In International Conference on Machine Learning (pp. 6105–6114). PMLR.
[28]. Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 1251–1258).
[29]. Han, K., Wang, Y., Tian, Q., Guo, J., Xu, C., & Xu, C. (2020). GhostNet: More features from cheap operations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 1580–1589).
[30]. Li, W., Li, S., & Liu, R. (2020, October 25–28). Channel shuffle reconstruction network for image compressive sensing. In 2020 IEEE International Conference on Image Processing (ICIP) (pp. 2880–2884). Institute of Electrical and Electronics Engineers.
[31]. Gao, X., Xu, L., Wang, F., et al. (2023). Multi-branch aware module with channel shuffle pixel-wise attention for lightweight image super-resolution. Multimedia Systems, 29(1), 289–303.
[32]. Lai, L., & Suda, N. (2018). Rethinking machine learning development and deployment for edge devices [Preprint]. arXiv. https://arxiv.org/abs/1806.07846
[33]. Han, L., Xiao, Z., & Li, Z. (2024). DTMM: Deploying TinyML models on extremely weak IoT devices with pruning [Preprint]. arXiv. https://arxiv.org/abs/2401.09068
Cite this article
Zhang,S. (2025). Review of Tiny Machine Learning Model Pruning Techniques on Resource-constrained Devices. Applied and Computational Engineering,145,29-36.
Data availability
The datasets used and/or analyzed during the current study will be available from the authors upon reasonable request.
Disclaimer/Publisher's Note
The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of EWA Publishing and/or the editor(s). EWA Publishing and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
About volume
Volume title: Proceedings of the 3rd International Conference on Software Engineering and Machine Learning
© 2024 by the author(s). Licensee EWA Publishing, Oxford, UK. This article is an open access article distributed under the terms and
conditions of the Creative Commons Attribution (CC BY) license. Authors who
publish this series agree to the following terms:
1. Authors retain copyright and grant the series right of first publication with the work simultaneously licensed under a Creative Commons
Attribution License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this
series.
2. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the series's published
version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial
publication in this series.
3. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and
during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See
Open access policy for details).