摘要
仿生模式识别神经网络(BPRNN)同传统BP、RBF神经网络相比具有更好的模式识别能力;训练样本库变更后网络的重新训练时间更小,但该网络构造过程中样本覆盖几何体参数的选择对网络识别率和复杂度有很大影响.本文通过引入蚁群算法来构造并优化网络参数,实验证明该算法法能较好的平衡网络性能和复杂度.
The pattern recognition capabilities of bionic pattern recognition neural networks（BPRNN） is better than BP or RBF neural networks;BPRNN consumes less time for re-training when the training sample set is changed.But during the construction process for BPRNN,the choice for the parameters of covering geometry has a great influence to the recognition rate or complexity of BPRNN.This paper proposed a method to construct and optimize network parameters by introducing ant colony algorithm.Experiment showed that the method can take better balance between performance and complexity of BPRNN.
出处
《闽江学院学报》
2013年第2期88-91,共4页
Journal of Minjiang University
关键词
蚁群算法
仿生模式识别
神经网络
ant colony algorithm
bionic pattern recognition
neural networks
