摘要
针对粒子群优化(PSO)算法搜索空间有限,容易陷入局部最优点的缺陷,提出一种以量子粒子群优化(QPSO)算法为基础的RBF神经网络训练算法,将RBF神经网络的参数组成一个多维向量,作为算法中的粒子进行进化,由此在可行解空间范围内搜索最优解。实例仿真表明,该学习算法相比于传统的学习算法计算简单,收敛速度快,并由于其算法模型的自身特性比基于PSO的学习算法具有更好的全局收敛性能。
Coping with such hmitations of Particle Swarm Optimization （PSO） algorithm as finite samphng space, being easy to run into local optima, a new Radial Basis Function Neural Network （ RBF NN） training method based on Quantumbehaved Particle Swarm Optimization （QPSO） algorithm was proposed. A multidimensional vector composed of RBF NN parameters was regarded as a particle in this algorithm to evolve. Then, the feasible sampling space was searched for the global optima. The simulation results show that this learning algorithm has easier computation and more rapid convergence compared with other traditional learning algorithms. And due to the characteristic of the algorithm model, its global convergence ability is better than the one based on PSO.
出处
《计算机应用》
CSCD
北大核心
2006年第8期1928-1931,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60474030)
关键词
粒子群优化算法
量子粒子群优化算法
径向基函数神经网络
Particle Swarm Optimization（PSO） algorithm
Quantum-behaved Particle Swarm Optimization（QPSO） algorithm
Radial Basis Function Neural Network （ RBF NN）
