卷积神经网络的损失最小训练后参数量化方法

doi:10.11959/j.issn.1000-436x.2022068

Abstract

Abstract:

To solve the problem that that no dataset is available for model quantization in data-sensitive scenarios, a model quantization method without using data sets was proposed.Firstly, according to the parameters of batch normalized layer and the distribution characteristics of image data, the simulated input data was obtained by error minimization method.Then, by studying the characteristics of data rounding, a factor dynamic rounding method based on loss minimization was proposed.Through quantitative experiments on classification models such as GhostNet and target detection models such as M2Det, the effectiveness of the proposed quantification method for image classification and target detection models was verified.The experimental results show that the proposed quantization method can reduce the model size by about 75%, effectively reduce the power loss and improve the computing efficiency while basically maintaining the accuracy of the original model.

Key words: convolutional neural network, batch normalization, simulation input data, dynamic rounding

CLC Number:

TP391

Fan ZHANG, Yun HUANG, Zizhuo FANG, Wei GUO. Lost-minimum post-training parameter quantization method for convolutional neural network[J]. Journal on Communications, 2022, 43(4): 114-122.

Figures/Tables 8

References 20

[1]	郭璠, 张泳祥, 唐琎 ,等. YOLOv3-A:基于注意力机制的交通标志检测网络[J]. 通信学报, 2021,42(1): 87-99.
	GUO F , ZHANG Y X , TANG J ,et al. YOLOv3-A:a traffic sign detection network based on attention mechanism[J]. Journal on Communications, 2021,42(1): 87-99.
[2]	黄志清, 曲志伟, 张吉 ,等. 基于深度强化学习的端到端无人驾驶决策[J]. 电子学报, 2020,48(9): 1711-1719.
	HUANG Z Q , QU Z W , ZHANG J ,et al. End-to-end autonomous driving decision based on deep reinforcement learning[J]. Acta Electronica Sinica, 2020,48(9): 1711-1719.
[3]	IOFFE S , SZEGEDY C . Batch normalization:accelerating deep network training by reducing internal covariate shift[C]// Proceedings of the 32nd International Conference on International Conference on Machine Learning.[S.l.]: JMLR.org, 2015: 448-456.
[4]	CHOUKROUN Y , KRAVCHIK E , YANG F ,et al. Low-bit quantization of neural networks for efficient inference[C]// Proceedings of 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW). Piscataway:IEEE Press, 2019: 3009-3018.
[5]	QIN H T , GONG R H , LIU X L ,et al. Forward and backward information retention for accurate binary neural networks[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 2247-2256.
[6]	ESSER S K , MCKINSTRY J L , BABLANI D ,et al. Learned step size quantization[J]. arxiv Preprint,arxiv:1902.08153, 2019.
[7]	NAGEL M , BAALEN M V , BLANKEVOORT T ,et al. Data-free quantization through weight equalization and bias correction[C]// Proceedings of 2019 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway:IEEE Press, 2019: 1325-1334.
[8]	LIU Y A , ZHANG W , WANG J . Zero-shot adversarial quantization[C]// Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2021: 1512-1521.
[9]	CAI Y H , YAO Z W , DONG Z ,et al. ZeroQ:a novel zero shot quantization framework[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 13166-13175.
[10]	NAGEL M , AMJAD R A , BAALEN M V ,et al. Up or down? adaptive rounding for post-training quantization[C]// Proceedings of 2020 International Conference on Machine Learning. New York:ACM Press, 2020: 7197-7206.
[11]	HE K M , ZHANG X Y , REN S Q ,et al. Deep residual learning for image recognition[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[12]	DENG J , DONG W , SOCHER R ,et al. ImageNet:a large-scale hierarchical image database[C]// Proceedings of 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2009: 248-255.
[13]	SZEGEDY C , VANHOUCKE V , IOFFE S ,et al. Rethinking the inception architecture for computer vision[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 2818-2826.
[14]	HAN D , YUN S , HEO B ,et al. Rethinking channel dimensions for efficient model design[C]// Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2021: 732-741.
[15]	SANDLER M , HOWARD A , ZHU M L ,et al. MobileNetV2:inverted residuals and linear bottlenecks[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 4510-4520.
[16]	RADOSAVOVIC I , KOSARAJU R P , GIRSHICK R ,et al. Designing network design spaces[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 10425-10433.
[17]	HAN K , WANG Y H , TIAN Q ,et al. GhostNet:more features from cheap operations[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2020: 1577-1586.
[18]	LIN T Y , MAIRE M , BELONGIE S J ,et al. Microsoft COCO:common objects in context[C]// Proceedings of 2014 European Conference on Computer Vision. Berlin:Springer, 2014: 740-755.
[19]	ZHANG S F , WEN L Y , BIAN X ,et al. Single-shot refinement neural network for object detection[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 4203-4212.
[20]	ZHAO Q J , SHENG T , WANG Y T ,et al. M2Det:a single-shot object detector based on multi-level feature pyramid network[C]// Proceedings of the AAAI Conference on Artificial Intelligence. New York:ACM Press, 2019: 9259-9266.

Metrics

Recommended 0

No Suggested Reading articles found!

实验次序	无数据激活量化方法			ZeroQ框架的激活量化方法
	模拟数据生成时间/μs	准确率		模拟数据生成时间/μs	准确率
	模拟数据生成时间/μs	top1	top5	模拟数据生成时间/μs	top1	top5
1	160 935	76.02%	92.88%	362 118	75.94%	92.86%
2	110 693	75.98%	92.84%	364 473	76.02%	92.89%
3	103 880	76.00%	92.84%	382 389	76.01%	92.8%
4	99 332	75.98%	92.80%	393 344	76.01%	92.8%
5	153 992	75.99%	92.84%	359 681	76.00%	92.8%
6	82 687	76.00%	92.83%	398 909	76.03%	92.83%
7	90 528	76.03%	92.88%	382 489	76.04%	92.86%
8	89 355	75.98%	92.84%	407 976	76.01%	92.85%
9	157 200	76.06%	92.87%	386 953	75.95%	92.82%
10	95 509	76.02%	92.88%	369 122	76.00%	92.86%
平均	114 411.1	76.01%	92.85%	380 745.4	76.00%	92.85%

模型	原始模型准确率		基于损失最小的无数据量化准确率			量化误差		模型大小
模型	top1	top5	实验次序	top1	top5	top1	top5	量化前/MB	量化后/MB	压缩率
			1	75.67%	92.73%	0.46%	0.13%
ResNet50	76.13%	92.86%	2	75.66%	92.73%	0.47%	0.13%	97.8	24.48	74.96%
			3	75.63%	92.74%	0.50%	0.12%
			平均	75.653%	92.733%	0.477%	0.127%
			1	78.49%	94.22%	0.34%	0.20%
Inception-v3	78.83%	94.42%	2	78.51%	94.22%	0.32%	0.20%	91.3	22.87	74.95%
			3	78.55%	94.21%	0.28%	0.21%
			平均	78.517%	94.217%	0.313%	0.203%
			1	82.28%	96.15%	0.35%	0.10%
RexNet3	82.63%	96.25%	2	82.21%	96.14%	0.42%	0.11%	132	33.07	74.95%
			3	82.21%	96.14%	0.42%	0.11%
			平均	82.233%	96.143%	0.397%	0.107%

模型	原始模型准确率		基于损失最小的无数据量化准确率			量化误差
模型	top1	top5	权重/位宽	top1	top5	top1	top5
MobileNetV2	73.03%	91.13%	W8/A8	71.96%	90.57%	1.07%	0.56%
			W16/A16	73.03%	91.14%	0%	-0.01%
RegNet	69.85%	89.37%	W8/A8	67.61%	88.30%	2.24%	1.07%
			W16/A16	69.81%	89.35%	0.04%	0.02%
GhostNet	73.98%	91.46%	W8/A8	73.02%	91.14%	0.96%	0.32%
			W16/A16	73.66%	91.39%	0.32%	0.07%

模型	原始模型平均准确率			量化模型平均准确率			量化误差
模型	AP_50:95	AP₅₀	AP₇₅	AP_50:95	AP₅₀	AP₇₅	AP_50:95	AP₅₀	AP₇₅
RefineDet_ResNet101_512	44.9%	67.1%	50.1%	43.9%	66.2%	49.0%	1.0%	0.9%	1.1%
M2Det_Vgg16_512	48.9%	68.2%	55.0%	48.3%	68.0%	54.5%	0.6%	0.2%	0.5%

Lost-minimum post-training parameter quantization method for convolutional neural network

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 8

References 20

Related Articles 15

Metrics

Recommended 0

[1]	Yurong LIAO, Haining WANG, Cunbao LIN, Yang LI, Yuqiang FANG, Shuyan NI. Research progress of deep learning-based object detection of optical remote sensing image [J]. Journal on Communications, 2022, 43(5): 190-203.
[2]	Zhengyu ZHU, Gengwang HOU, Chongwen HUANG, Gangcan SUN, Wanming HAO, Jing LIANG. Systems resource allocation algorithm for RIS-assisted D2D secure communication based on parallel CNN [J]. Journal on Communications, 2022, 43(3): 172-179.
[3]	Hongyan WANG, Libin ZHANG, Guoqiang CHEN, Zumin WANG, Zhiyuan GUAN. Approach of target tracking combining particle filter and metric learning [J]. Journal on Communications, 2021, 42(5): 98-110.
[4]	Nianwen SI, Wenlin ZHANG, Dan QU, Heyu CHANG, Shengxiang LI, Tong NIU. Generalized Grad-CAM attacking method based on adversarial patch [J]. Journal on Communications, 2021, 42(3): 23-35.
[5]	Hongmin GAO, Xueying CAO, Zhonghao CHEN, Zaijun HUA, Chenming LI, Yue CHEN. Hyperspectral image classification method based on multi-scale proximal feature concatenate network [J]. Journal on Communications, 2021, 42(2): 92-102.
[6]	Jun YANG,Jisheng DANG. Semantic segmentation of 3D point cloud based on contextual attention CNN [J]. Journal on Communications, 2020, 41(7): 195-203.
[7]	Hongmin GAO,Xueying CAO,Yao YANG,Zaijun HUA,Chenming LI. Application of bilateral fusion model based on CNN in hyperspectral image classification [J]. Journal on Communications, 2020, 41(11): 132-140.
[8]	Meng ZHANG,Haoliang SUN,Peng YANG. Identification of DNS covert channel based on improved convolutional neural network [J]. Journal on Communications, 2020, 41(1): 169-179.
[9]	Xin ZHOU,Xiaoxin HE,Changwen ZHENG. Radio signal recognition based on image deep learning [J]. Journal on Communications, 2019, 40(7): 114-125.
[10]	Jia LI,Xiaochun YUN,Shuhao LI,Yongzheng ZHANG,Jiang XIE,Fang FANG. HTTP malicious traffic detection method based on hybrid structure deep neural network [J]. Journal on Communications, 2019, 40(1): 24-33.
[11]	Linhui LI,Bo QIAN,Jing LIAN,Weina ZHENG,Yafu ZHOU. Study on traffic scene semantic segmentation method based on convolutional neural network [J]. Journal on Communications, 2018, 39(4): 123-130.
[12]	Hengjun WANG,Nianwen SI,Yulong SONG,Yidong SHAN. Neural network model for dependency parsing incorporating global vector feature [J]. Journal on Communications, 2018, 39(2): 53-64.
[13]	Wanliang WANG,Zhuorong LI. Advances in generative adversarial network [J]. Journal on Communications, 2018, 39(2): 135-148.
[14]	Yong WANG,Huiyi ZHOU,Hao FENG,Miao YE,Wenlong KE. Network traffic classification method basing on CNN [J]. Journal on Communications, 2018, 39(1): 14-23.
[15]	Bing-lin ZHAO,Xi MENG,Jin HAN,Jing WANG,Fu-dong LIU. Homology analysis of malware based on graph [J]. Journal on Communications, 2017, 38(Z2): 86-93.