摘要
针对即使是全局收敛的变型标准遗传算法VCGA[1]有时也会发生收敛速度变慢的问题,提出了VCGA和最陡下降法相结合的混合法HVCSDA.该方法增强了VCGA在接近全局最优解时的微调能力.还将HVCSDA推广到一类修正的VCGA上.仿真实例表明了HVCSDA及其推广能有效地提高收敛速度.对30城市TSP的仿真结果为6.822,要好于用TABU得到的6.99的结果[6].
This paper first surveys the structure, types, approximation theory, and training methods of the RBF neural network, then analyzes the advantages and problems. At the same time, some typical applications in control have been discussed. At last the new trends of the applications in control are pointed out.
出处
《信息与控制》
CSCD
北大核心
1997年第4期266-271,共6页
Information and Control
基金
国家自然科学基金
关键词
VCGA
收敛速度
混合法
遗传算法
RBF neural networks, universal approximation, best approximation, predictive control, internal model control, soft sensor, optimization of operation