摘要
研究粒子群优化算法。传统的粒子群算法采用实数编码,收敛速度慢。为了提高收敛速度,提出了一种混沌编码的粒子群优化算法。混沌编码作为一种全新的数学编码方式,更能准确地表达编码对象的多样性,将混沌编码应用到粒子群优化算法中,使算法在初期的搜索区域更大,更快找到全局最优解。把混沌编码的粒子群算法与BP算法相结合用来优化神经网络。利用混沌编码的粒子群算法快速找到全局最优位置的邻域,然后再用BP算法进行局部寻优,收敛到全局最优位置。仿真结果证明混沌编码的粒子群神经网络比实数编码的粒子群神经网络分类收敛速度更快,验证了算法的有效性。
In order to improve convergence speed, this paper presented a chaos encoded particle swarm optimiza- tion new mathematical encoding, wieh can accurately express the diversity of the encoded object. Chaotic encoding was applied to the particle swarm algorithm. The algorithm searches a larger area at initial and finds the global optimal solution more quickly. Chaotic coded particle swarm algorithm and BP algorithms were combined to optimize the neu- ral network to solve the classification problem. Chaotic coding particle swarm algorithm can quickly find the location of the global optimum in the neighborhood. Then BP algorithm carries out local search and finds the global optimum position. The simulation results show that convergence speed of the neural network classification based on chaotic co- ded particle swarm is faster than the one based on real-coded particle swarm. The paper verifies the effectiveness of the algorithm.
出处
《计算机仿真》
CSCD
北大核心
2013年第3期299-302,共4页
Computer Simulation
基金
江西省教育厅科技项目(GJJ09253)
关键词
混沌编码
粒子群优化
神经网络
收敛速度
Chaotic coding
Particle swarm optimization (PSO)
Neural network (NN)
Convergence speed