摘要
本文采用遗传编程(Genetic Programming,下用GP表示)的方法,提出一种对神经网络如何同时优化它的权值和结构(包括层数、每层的神经元个数以及神经元之间的连接关系)的新思想。遗传编程(Genetic Programming)是遗传算法(Genetic Alogrithms)的扩展,利用具有可变长度的LISP符号表达式表示群体中的个体,是基于达尔文(Darwin)的进化论和遗传基因学原理的一种新兴的搜索寻优技术。本文采用这种方法设计了一个神经网络,成功解决了一位加法器问题。
This paper shows how to find both the weights and architecture for a neural network (ineluding the number of layers, the number of processing elements per layer, and the connectivity between processing elements). This is accomplished using the new 'genetic programming' paradigrn. It's a recently developed extension to genetic algorithm which breeds a population of LISP symbotic expressions of varying size and shape until the desired performance by the network is successfully evolved. The new method is applied to the problem of generating a neural network for the one-bit adder.
出处
《中山大学研究生学刊（自然科学与医学版）》
2000年第4期1-6,共6页
Journal of the Graduates Sun YAT-SEN University(Natural Sciences.Medicine)
