摘要
相比于传统的基于梯度的前馈神经网络,随机前馈神经网络具有更好的逼近能力和泛化学习能力,被广泛用于分类等问题中,然而其网络参数完全随机,在实际应用中存在性能不稳定、不可靠的隐患。为此,基于人类学习优化算法提出了一种改进的选择性进化随机网络方法(Improved Selective Evolutionary Random Network,ISERN),协同进行特征选择和网络结构优化以提高网络性能,某远洋船舶海水淡化系统的故障诊断仿真结果表明ISERN方法与其他方法相比具有更好的故障诊断性能,体现出其有效性和优异性。
Aiming at the unstable performance of random feedforward neural network,an improved evolutionary random network(ISERN) based on the human learning optimization(HLO) algorithm is proposed,in which the structure of the random feedforward neural network is optimized and the feature selection is carried out cooperatively and simultaneously as a wrapper.The simulation results of fault diagnosis of the ocean-going ship desalination system show the proposed ISERN method has a satisfactory classification ability.
出处
《工业控制计算机》
2021年第11期93-95,97,共4页
Industrial Control Computer
关键词
人类学习优化算法
随机前馈神经网络
进化随机网络
故障诊断
human learning optimization
random feedforward neural network
evolutionary random network
fault diagnosis
