Abstract
Quantization allows accelerating neural networks significantly, especially for mobile processors. Existing quantization methods require either training neural network from scratch or gives significant accuracy drop for the quantized model. Low bits quantization (e.g., 4- or 6-bit) task is a much more resource consumptive problem in comparison with 8-bit quantization, it requires a significant amount of labeled training data. We propose a new low-bit quantization method for mobile neural network architectures that doesn’t require training from scratch and a big amount of train labeled data and allows to avoid significant accuracy drop.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Jacob, B., et al.: Quantization and training of neural networks for efficient integer-arithmetic-only inference. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2704–2713 (2018)
Goncharenko, A., Denisov, A., Alyamkin, S., Terentev, E.: Fast adjustable threshold for uniform neural network quantization. Int. J. Comput. Inf. Eng. 13(9), 495–499 (2019)
Banner, R., Nahshan, Y., Hoffer, E., Soudry, D.: Post-training 4-bit quantization of convolution networks for rapid-deployment. arXiv preprint arXiv:1810.05723 (2018)
Deng, J., et al.: ImageNet: a large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp. 248–255. IEEE, June 2009
Tan, M., Le, Q.: EfficientNet: rethinking model scaling for convolutional neural networks. In: International Conference on Machine Learning, pp. 6105–6114. PMLR, May 2019
Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)
Kullback, S.: Information Theory and Statistics. John Riley and sons. Inc., New York (1959)
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., Chen, L.C.: MobileNetV2: inverted residuals and linear bottlenecks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4510–4520 (2018)
He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016)
Krishnamoorthi, R.: Quantizing deep convolutional networks for efficient inference: a whitepaper. arXiv preprint arXiv:1806.08342 (2018)
Hinton, G., Vinyals, O., Dean, J.: Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531 (2015)
Migacz, S.: GPU Technology Conference, pp. 10–30 (2017)
Author information
Authors and Affiliations
Corresponding authors
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 Springer Nature Switzerland AG
About this paper
Cite this paper
Chudakov, D., Alyamkin, S., Goncharenko, A., Denisov, A. (2021). Iterative Adaptation to Quantization Noise. In: Rojas, I., Joya, G., Català, A. (eds) Advances in Computational Intelligence. IWANN 2021. Lecture Notes in Computer Science(), vol 12861. Springer, Cham. https://doi.org/10.1007/978-3-030-85030-2_25
Download citation
DOI: https://doi.org/10.1007/978-3-030-85030-2_25
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-85029-6
Online ISBN: 978-3-030-85030-2
eBook Packages: Computer ScienceComputer Science (R0)