摘要
设计卷积升级网络结构,引入PSO算法减小了误差的反向传播,避免了滞后误差与图像的过拟合,提高了收敛速度。将该方法应用到数据集HCL2000和MNIST上,并进行了与WCNN、MLP-CNN、SVM-ELM的实验对比,证明了改进算法的正确性。
A convolution upgrade network structure is designed, and then PSO algorithm is introduced to reduce the error back propagation. Accordingly, the over-fit between delayed error and image is avoided for improving convergent speed. The method is applied to data set HCL2000 and MNIST, and experiments compared with WCNN, MLP-CNN and SVM-ELM are carried.
作者
王金哲
王泽儒
王红梅
WAN Jinzhe;WANG Zeru;WANG Hongmei(School of Computer Science&Engineering,Changchun University of Technology,Changchun 130012,China)
出处
《长春工业大学学报》
CAS
2019年第1期26-30,共5页
Journal of Changchun University of Technology
基金
吉林省科技厅科技发展计划基金资助项目(20160203010GX)
