计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (9): 162-167.DOI: 10.3778/j.issn.1002-8331.1901-0122

• 模式识别与人工智能 • 上一篇    下一篇

局部感受野的宽度学习算法及其应用

李国强,徐立庄   

  1. 燕山大学 电气工程学院,河北 秦皇岛 066004
  • 出版日期:2020-05-01 发布日期:2020-04-29

Application of Local Receptive Field Based Broad Learning System

LI Guoqiang, XU Lizhuang   

  1. School of Electrical Engineering, Yanshan University, Qinhuangdao, Hebei 066004, China
  • Online:2020-05-01 Published:2020-04-29

摘要:

为了更快且更准确地对图像进行识别,提出了基于局部感受野的宽度学习算法(Local Receptive Field based Broad Learning System,BLS-LRF),该方法以宽度学习网(Broad Learning System,BLS)为基础模型,与局部感受野(LRF)的思想相结合,从局部特征和全局特征两方面对图像进行特征提取。采用两种图像数据集对网络进行研究,将研究结果和许多传统神经网络进行对比,结果表明BLS-LRF网络的测试精度不仅超过了传统网络的测试精度,而且训练过程所需要的时间有了很大程度的缩短。

关键词: 宽度学习网, 局部感受野, 神经网络, 图像分类

Abstract:

In order to identify images more quickly and accurately, this paper puts forward the Local Receptive Field based Broad Learning System(BLS-LRF) algorithm, this method is based on the Broad Learning System(BLS) model and combines with the idea of Local Receptive Field(LRF) to extract the image features from two aspects:local feature and global feature. Two kinds of image data sets are used to study the network, and the results are compared with many traditional neural networks. The results show that the test accuracy of the BLS-LRF network not only exceeds the test accuracy of the traditional network, but also greatly shortens the time required for the training process.

Key words: broad learning system, Local Receptive Field(LRF), neural network, image classification