基于模块电路结构的BP神经网络及其应用研究
A MODULAR STRUCTURE BASED BP NEURAL NETWORK AND ITS APPLICATION
-
摘要: 本文采用一种简化的BP(Back Propagation)神经网络硬件模块实现方法。该方法利用全电流模式电路组成神经元模块,再用若干模块构成简化的BP神经网络。所提出的模块结构网络系统具有在线学习和在线权值存储能力,且可应用于实现编、解码和二维图像识别。文中提供了PSPICE和高级语言计算机仿真结果。Abstract: This paper presents a hardware implementation approach for realizing simple BP (Backward Propagation) neural network. The full current-mode analog circuits are used to form neuron modules, and a simple BP network is build using basic modules. This network has the property of on-chip learning and on-chip weight storing, and it can be used for coding, decoding and two-dimensional image recognition. Simuiation results with PSPICE and high-level languages are given.
-
Bibyk S, et al. Current-mode neural network building blocks for analog MOS VLSI. IEEE ISCAS,[2]Helsinki: 1990, 3283-3285.[3]Mead C. Analog VLSI and Nearal Networks. Reading, Meassachusetts: Addison Wesley, 1989. [3] 庞维珍, 任鲁涌, 等. 在片学习及权值刷新神经网络硬件实现方法的研究. 天津大学学报. 1996, 29(6): 821-827.[4]Salam F, Choi M. An all-MOS analog feedforward neural circuit with learning. IEEE ISCAS, Helsinki: 1990, 2508-2511. -
计量
- 文章访问数: 2398
- HTML全文浏览量: 109
- PDF下载量: 411
- 被引次数: 0
下载:
