分组合作多智能体算法优化BP神经网络的权值以及阈值.pdfVIP

BP （南京邮电大学自动化学院，江苏 南京，210003） 本文提出了一种基于分组合作的多智能体算法用来优化BP神经网络的学习算法。该算法是要设计出一 个多智能网格，把这些网格分成相互独立的若干个小组，首先小组内部的各个智能体合作，然后选出最优的 智能体随机与网格中的智能体合作，每个智能体都有一定的概率自我变异。先通过该算法来训练参数达到一 定的要求，然后在通过BP神经网络算法来训练。该算法极大的提高了BP神经网络的收敛速度。通过多项式逼 近函数，证明了该算法非常的有效。 多智能体；遗传算法；BP神经网络；函数逼近 TP183 A Group cooperation Multi-agent optimize the weight and threshold of BP neural network (College of Automation, Nanjing University of Posts and Telecommunications,Nanjing,Jiangsu,210003) Abstract: This paper presents an optimization algorithm of BP neutral network based on learning of multi-agents group cooperation. This algorithm is to design a multi-grid, the grid is divided into those of a number of independent groups, first of all within the group agent cooperation, then select the best agent in the random and grid-agent co- operation each agent has a certain probability of self-mutation. First through the algorithm to train the parameters meet certain requirements, and then through the BP neural network algorithm to train. The algorithm greatly improves the convergence speed of BP neural network. Through the polynomial approximation function, show that the algorithm is very effective. Keywords: multi-agent;genetic algorithm;BP neural network; function approximation 2222