摘要
在对仓虫分类识别过程中,为了改善因采用BP神经网络产生的由于训练时间长和易于陷入局部极小点,而导致效率和分类的准确性较低的情况,对粒子群优化算法进行了研究,并把这种算法运用到神经网络学习训练中。实验表明,将基于粒子群优化的神经网络算法应用到仓虫分类中,从训练时间、识别率上得到了较大的改善,而且算法易于实现,且能更快地收敛于全局最优解。
Using back-propagation neural network to recognise and classify grain pests is inefficient and has low accuracy in classification,because it spends too long time in training and falls easily into local minimum.In order to improve this,Particle swarm optimisation（PSO） was studied in this paper,and was applied to training neural network learning.Experimental result showed that applying PSO-based neural network algorithm in recognising and classifying grain pests can obviously improve the recognition rate and reduce the training time,the algorithm is easy to implement,and converges to global optimal solution quickly.
出处
《计算机应用与软件》
CSCD
2010年第1期228-230,共3页
Computer Applications and Software
关键词
粒子群优化算法
神经网络
仓虫
特征提取
Particle swarm optimisation Neural network Grain pests Feature extraction
