Neuron-by-Neuron Quantization for Efficient Low-Bit QNN Training
Mathematics · 29.04.2023 · 11 (9) · 2112
DOI: 10.3390/math11092112 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics, online erschienen am 29.04.2023. Verfasst von Artem Sher und 4 weiteren Autor:innen. Die Arbeit wurde bisher 13-mal zitiert (Crossref) und nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Science and Higher Education of the Russian Federation.
Abstract (Originalsprache)
Quantized neural networks (QNNs) are widely used to achieve computationally efficient solutions to recognition problems. Overall, eight-bit QNNs have almost the same accuracy as full-precision networks, but working several times faster. However, the networks with lower quantization levels demonstrate inferior accuracy in comparison to their classical analogs. To solve this issue, a number of quantization-aware training (QAT) approaches were proposed. In this paper, we study QAT approaches for two- to eight-bit linear quantization schemes and propose a new combined QAT approach: neuron-by-neuron quantization with straight-through estimator (STE) gradient forwarding. It is suitable for quantizations with two- to eight-bit widths and eliminates significant accuracy drops during training, which results in better accuracy of the final QNN. We experimentally evaluate our approach on CIFAR-10 and ImageNet classification and show that it is comparable to other approaches for four to eight bits and outperforms some of them for two to three bits while being easier to implement. For example, the proposed approach to three-bit quantization of the CIFAR-10 dataset results in 73.2% accuracy, while baseline direct and layer-by-layer result in 71.4% and 67.2% accuracy, respectively. The results for two-bit quantization for ResNet18 on the ImageNet dataset are 63.69% for our approach and 61.55% for the direct baseline.
Förderung
- Ministry of Science and Higher Education of the Russian Federation
Zitieren
APA
Sher, A., Trusov, A., Limonova, E., Nikolaev, D., & Arlazarov, V. V. (2023). Neuron-by-Neuron Quantization for Efficient Low-Bit QNN Training. Mathematics, 11(9), 2112. https://doi.org/10.3390/math11092112
Vancouver
Sher A, Trusov A, Limonova E, Nikolaev D, Arlazarov VV. Neuron-by-Neuron Quantization for Efficient Low-Bit QNN Training. Mathematics. 2023;11(9):2112. doi:10.3390/math11092112
BibTeX
@article{Sher2023,
author = {Sher, Artem and Trusov, Anton and Limonova, Elena and Nikolaev, Dmitry and Arlazarov, Vladimir V.},
title = {Neuron-by-Neuron Quantization for Efficient Low-Bit QNN Training},
journal = {Mathematics},
year = {2023},
volume = {11},
number = {9},
pages = {2112},
doi = {10.3390/math11092112}
}
EUVIMED