光学学报, 2002, 22 (1): 62, 网络出版: 2006-08-08
洗牌型光电混合神经网络实验系统
The Implementation of a Perfect Shuffle Type of Optoelectronic Neural Network System
光学神经网络 洗牌 图样间联想 光电混合神经网络 optical neural networks perfect shuffle interpattern association hybrid opto-electrical neural network
摘要
在研究洗牌网局部互连光学实现问题的基础上,采用洗牌型图样间联想模型建立了具有完整的加权互连、求和、非线性处理及反馈功能的光电混合神经网络实验系统,进行了8×8数字样本的光学联想识别.实验结果证实了洗牌网理论的可行性和洗牌型图样间联想模型光学实现的优越性.
Abstract
Based on the study of optical implementation of the perfect shuffle local interconnecting, a hybrid opto-electrical perfect shuffle type of neural network experiment system is constructed with full functions of weighted interconnecting and summing, nonlinear processing and feeding back. Optical associative recognizing experiments for numeral patterns in 8×8 pixels are proceeded employing the PS-IPA algorithm. The experimental results show a good agreement with theoretical simulations and the feasibility and superiority of the shuffle network theory is then confirmed. And large-scale perfect shuffle type of practical ONNs with 10 4 neurons are expected utilizing smart pixel array device.
许锐, 黄达诠, 李志能. 洗牌型光电混合神经网络实验系统[J]. 光学学报, 2002, 22(1): 62. 许锐, 黄达诠, 李志能. The Implementation of a Perfect Shuffle Type of Optoelectronic Neural Network System[J]. Acta Optica Sinica, 2002, 22(1): 62.