多感官群集智能算法及其在前向神经网络训练方面的应用

针对连续域函数优化问题,提出了一种新的全局极大值搜索方法--多感官群集智能算法(multi-sense swarm intelligence algorithm,MSA). 受鱼群算法(artificial fish-swarm algorithm,AFA)和FS算法(free search algorithm,FSA)的启发,MSA的搜索机制将大范围勘察和小范围精确搜索相结合,个体在使用视觉信息快速逼近局部较优解的同时,利用嗅觉信息避免群体过于集中并引导个体向全局较优解方向移动. 仿真结果证明:MSA鲁棒性较强,全局收敛性好,收敛速度较快,收敛精度较高. 最后,将该方法应用于前向神经网络训练,结果表明满足应用要求.

Multi-sense swarm intelligence algorithm and its application in feed-forward neural networks training

A novel method for global optimization,multi-sense swarm intelligence algorithm(MSA),was presented to solve continuous function optimization problems.Inspired by the artificial fish-swarm algorithm(AFA) and the FS algorithm(free search algorithm,FSA),the search mechanism of MSA combined large scale exploration and local precise search;even more,in this algorithm,the unit employed both visual information for quick approaching to local optimization solution and pheromone information to avoid overcrowding and to guide itself to global solution. Simulation shows that MSA has strong robustness,good global convergence,quick convergence speed and high convergence accuracy.At last,MSA was applied to feed-forward neural network training.The result shows that this algorithm is fit for the application.