计算机与现代化
計算機與現代化
계산궤여현대화
Computer and Modernization
2015年
9期
57-59,65
,共4页
蛙跳算法%粒子群算法%BP神经网络
蛙跳算法%粒子群算法%BP神經網絡
와도산법%입자군산법%BP신경망락
shuffled frog leaping algorithm%particle swarm optimization algorithm%BP neural network
针对BP神经网络算法计算量复杂、收敛速度缓慢等缺点，提出一种基于启发式算法的BP神经网络权值和阈值的迭代方法。该方法结合蛙跳粒子群可控参数少、收敛速度快等特点，将神经网络权值和阈值作为粒子，通过粒子更新来实现BP神经网络训练。实验结果表明，该算法的精度可在1.5342e-03左右。
針對BP神經網絡算法計算量複雜、收斂速度緩慢等缺點，提齣一種基于啟髮式算法的BP神經網絡權值和閾值的迭代方法。該方法結閤蛙跳粒子群可控參數少、收斂速度快等特點，將神經網絡權值和閾值作為粒子，通過粒子更新來實現BP神經網絡訓練。實驗結果錶明，該算法的精度可在1.5342e-03左右。
침대BP신경망락산법계산량복잡、수렴속도완만등결점，제출일충기우계발식산법적BP신경망락권치화역치적질대방법。해방법결합와도입자군가공삼수소、수렴속도쾌등특점，장신경망락권치화역치작위입자，통과입자경신래실현BP신경망락훈련。실험결과표명，해산법적정도가재1.5342e-03좌우。
In order to solve the problems that the calculation of BP neural network is very complex and its convergence rate is slow, an iterative method of weight and threshold of the BP neural network based on the heuristic algorithm is put forward. This method combined with two advantages of the Frog Leaping Particle Swarm, in which, one is less controllable parameter than nor-mal ways and the other one is the fast convergence speed. In essence, the weight and the threshold of neural network can be seen as particles. BP neural network was trained by particle updating and the accuracy of the algorithm is about 1. 5342e-03.
