【作者】 汪荣贵； 李守毅； 孙见青；

【Author】 Wang Ronggui, Li Shouyi, Sun Jianqing(School of Computer and Information, Heifei University of Technology, Hefei 230009,China)

【机构】 合肥工业大学计算机与信息学院；

【摘要】 量子粒子群优化算法(QPSO)是一种基于群体的随机优化技术。与标准的粒子群优化算法(PSO)相比,QPSO更具全局搜索能力并具有较少的控制参数等优点。本文以QPSO技术取代传统BP神经网络学习算法中的梯度下降法,使改进后的神经网络具有良好的全局收敛性。将改进后的学习算法应用于人脸检测,实验结果表明该算法在此应用上是有效的。更多还原

【Abstract】 Quantum-behaved Particle Swarm Optimization (QPSO) algorithm is a population-based stochastic optimization technique, which outperforms traditional PSOs in search ability as well as having less parameter to control. In this paper, QPSO technique is introduced into BP neural network to instead adopting gradient descent method in BP learning algorithm. Due to the characteristic of the QPSO algorithm, the problems of traditional BP learning algorithm can be avoided, such as easily converging to local minimum. Then we adopt the new learning algorithm to training a neural network for face detection, and the experiment results testify its efficiency in this paper.更多还原