DSConv: Efficient Convolution Operator

被引:69
作者
do Nascimento, Marcelo Gennari [1 ]
Fawcett, Roger [2 ]
Prisacariu, Victor Adrian [1 ]
机构
[1] Univ Oxford, Act Vis Lab, Oxford, England
[2] Intel Corp, Santa Clara, CA 95051 USA
来源
2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019) | 2019年
关键词
D O I
10.1109/ICCV.2019.00525
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Quantization is a popular way of increasing the speed and lowering the memory usage of Convolution Neural Networks (CNNs). When labelled training data is available, network weights and activations have successfully been quantized down to 1-bit. The same cannot be said about the scenario when labelled training data is not available, e.g. when quantizing a pre-trained model, where current approaches show, at best, no loss of accuracy at 8-bit quantizations. We introduce DSConv, a flexible quantized convolution operator that replaces single-precision operations with their far less expensive integer counterparts, while maintaining the probability distributions over both the kernel weights and the outputs. We test our model as a plug-and-play replacement for standard convolution on most popular neural network architectures, ResNet, DenseNet, GoogLeNet, AlexNet and VGG-Net and demonstrate state-of-the-art results, with less than 1% loss of accuracy, without retraining, using only 4-bit quantization. We also show how a distillation-based adaptation stage with unlabelled data can improve results even further.
引用
收藏
页码:5147 / 5156
页数:10
相关论文
共 40 条
[1]  
Abadi M, 2016, PROCEEDINGS OF OSDI'16: 12TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, P265
[2]  
[Anonymous], 2015, BinaryConnect: Training Deep Neural Networks with binary weights during propagations
[3]  
[Anonymous], 2018, TRAINING INFERENCE I
[4]  
[Anonymous], 2013, CORR
[5]  
[Anonymous], 2015, Arxiv.Org, DOI DOI 10.3389/FPSYG.2013.00124
[6]  
[Anonymous], 2015, High Performance Hardware for Machine Learning
[7]  
[Anonymous], 2015, PROCIEEE CONFCOMPUT, DOI DOI 10.1109/CVPR.2015.7298594
[8]  
Ba LJ, 2014, ADV NEUR IN, V27
[9]  
Banner Ron, 2018, POSTTRAINING 4 BIT Q
[10]  
Cai Z, 2017, WOODH PUBL SER BIOM, P171, DOI 10.1016/B978-0-08-100383-1.00010-2