一类神经网络多层感知器的全硬件实现

Hardware Implementation of the Multilayer Perceptron for a Neural Network

针对模式分类问题，给出了一种神经网络多层感知器数字电路硬件实现方法，该方法在硬件电路中不含有数字乘法器，从而有助于克服目前神经网络数字器件难以单片集成的缺限．采用该方法研制出了具体的硬件线路板，应用于一实际的分类问题。

The self learning function of the multilayer perceptron for an artificial neural network can be easily realized by dynamic change of the weights, but it is very difficult to bring about the change in hardware implementation. This paper presents a new method for accomplishing digital circuit implementation of the multilayer perceptron, which is especially useful for pattern classification. As there is no digital multiplier in the hardware circuit board, the difficulty in single chip integration can be overcome. A perceptron digital circuit for solving actual class ification has been successfully developed.