论文部分内容阅读
为了提高神经网络的泛化性,对输入信号进行频率分解.频率分解相对提升了子频带的信息致密性,覆盖全频域的子频带,也保证了信息的遍历性.高致密性和遍历性有助于提高神经网络的泛化性.频率分解由盲动粒子群优化算法自动完成,粒子群算法和通常的神经网络算法都用迭代计算,但计算需耗费较长时间,而采用一次就完成学习的极速学习神经网络可以节省计算时间.仿真结果表明,该神经网络泛化性好、精度高能满足一般工程应用.
In order to improve the generalization of the neural network, the input signal is decomposed by frequency, which improves the information compactness of subbands and covers the subbands in the whole frequency domain, and ensures the ergodicity of information. High density and ergodicity Help to improve the generalization of neural network.Frequency decomposition is done automatically by blindly moving particle swarm optimization algorithm, particle swarm optimization algorithm and the usual neural network algorithm are used iterative calculation, but the calculation takes a long time, and use once to complete the learning Speed learning neural network can save computing time.The simulation results show that the neural network has good generalization and high precision to meet the general engineering application.