多宇宙并行量子遗传神经网络人脸识别算法研究
|
摘要
针对传统遗传算法交叉、变异过程过于繁琐和神经网络在极值判断及收敛速度受限等问题,提出了一种并行的量子遗传算法优化神经网络权值的算法.首先引入了量子计算的概念,在量子计算的过程中使用量子旋门实现染色体的训练,然后引入量子交叉克服了早熟收敛现象,避免了遗传算法中繁琐的交叉、变异过程.最后设计实现了并行的卷积神经网络,使用并行量子遗传算法优化了卷积神经网络权值,实现了并行量子遗传神经网络人脸识别系统.实验结果表明,相对于原来的遗传算法,该算法在鲁棒性和实验速度上都有明显的提高.
In order to solve the problem that the process of cross and mutation in traditional genetic algorithm is too cumbersome,and the extreme value judgment and the convergence rate is limited,a parallel quantum genetic algorithm( QGA) is proposed to optimize the weights of the neural network. The concept of quantum computing is firstly introduced. In the process of quantum computation,the quantum rotation gate is used to train the chromosomes. Then the quantum cross is introduced to overcome the precocious convergence and to avoid the cumbersome cross and mutation process in the genetic algorithm. Finally,the parallel convolution neural network is designed and implemented. The parallel quantum genetic algorithm is used to optimize the weights of the convolution neural network,and a facial recognition system based on parallel quantum genetic neural network is realized. Experimental results show that compared with the original genetic algorithm, the quantum genetic neural network algorithm has obvious improvements in terms of robustness and processing speed.
In order to solve the problem that the process of cross and mutation in traditional genetic algorithm is too cumbersome,and the extreme value judgment and the convergence rate is limited,a parallel quantum genetic algorithm( QGA) is proposed to optimize the weights of the neural network. The concept of quantum computing is firstly introduced. In the process of quantum computation,the quantum rotation gate is used to train the chromosomes. Then the quantum cross is introduced to overcome the precocious convergence and to avoid the cumbersome cross and mutation process in the genetic algorithm. Finally,the parallel convolution neural network is designed and implemented. The parallel quantum genetic algorithm is used to optimize the weights of the convolution neural network,and a facial recognition system based on parallel quantum genetic neural network is realized. Experimental results show that compared with the original genetic algorithm, the quantum genetic neural network algorithm has obvious improvements in terms of robustness and processing speed.
引文
[1] Zhang X F,Sui G F,Zheng R,et al. An improved quantum genetic algorithm of quantum revolving gate[J]. Computer Engineering,2013,39(4):234-238.
[2] Fu D S,Zhang R. Research on an improved quantum genetic algorithms[J]. Computer Simulation,2013,30(12):321-325.
[3] Yang J A,Zhuang Z Q,Shi L. Multi-universe parallel genetic algorithms[J]. Electronic Journal,2004,32(6):923-928.
[4] Simonyan K,Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science,arXiv preprintarXiv:1409. 1556,2014.
[5] Gong D W,Sun X Y. Upper bound of crossover and mutation probability of genetic algorithm based on pattern theorem[J]. Control and Decision-Making,2004,19(5):554-556.
[6] Tan C C,Eswaran C. Reconstruction and recognition of face and digit images using autoencoders[J]. Neural Comput&Applic,2010,19(7):1069-1079.
[7] Alyüz N,Gokberk B,Akarun L. Regional registration for expression resistant 3-D face recognition[J]. IEEE Transactions on Information Forensics and Security,2010,5(3):425-440.
[8] Hsieh P,Tung P. Shadow compensation based on facial symmetry and image average for robust face recognition[J].Neurocomputing,2010,73(13/15):2708-2717.
[9] Wang L M,Qiao L L,Wei L J. An optimal learning method of convolutional neural network based on genetic algorithms[J].Computer Engineering and Design,2017,38(7):1946-1950.
[10] Yang W,Jin L,Tao D,et al. DropSample:a new training method to enhance deep convolutional neural networks for large-scale unconstrained handwritten chinese character recognition[J].Pattern Recognition,2015,58(4):190-203.
[2] Fu D S,Zhang R. Research on an improved quantum genetic algorithms[J]. Computer Simulation,2013,30(12):321-325.
[3] Yang J A,Zhuang Z Q,Shi L. Multi-universe parallel genetic algorithms[J]. Electronic Journal,2004,32(6):923-928.
[4] Simonyan K,Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science,arXiv preprintarXiv:1409. 1556,2014.
[5] Gong D W,Sun X Y. Upper bound of crossover and mutation probability of genetic algorithm based on pattern theorem[J]. Control and Decision-Making,2004,19(5):554-556.
[6] Tan C C,Eswaran C. Reconstruction and recognition of face and digit images using autoencoders[J]. Neural Comput&Applic,2010,19(7):1069-1079.
[7] Alyüz N,Gokberk B,Akarun L. Regional registration for expression resistant 3-D face recognition[J]. IEEE Transactions on Information Forensics and Security,2010,5(3):425-440.
[8] Hsieh P,Tung P. Shadow compensation based on facial symmetry and image average for robust face recognition[J].Neurocomputing,2010,73(13/15):2708-2717.
[9] Wang L M,Qiao L L,Wei L J. An optimal learning method of convolutional neural network based on genetic algorithms[J].Computer Engineering and Design,2017,38(7):1946-1950.
[10] Yang W,Jin L,Tao D,et al. DropSample:a new training method to enhance deep convolutional neural networks for large-scale unconstrained handwritten chinese character recognition[J].Pattern Recognition,2015,58(4):190-203.