基于Mackey-Glass混沌储备池计算的多波形识别

摘要

人工神经网络在模式识别和智能计算等领域具有重要的应用,但目前的人工神经网络存在结构复杂、训练耗时、占用大量的硬件资源等缺陷.设计了一种简单的基于 Mackey-Glass非线性方程的混沌储备池计算系统,它把大量不同时刻的混沌状态当作虚拟节点,从而简化了系统结构.利用系统分别实现了3种波形和对应3种不同频率的共9个不同波的识别,分析了储备池计算系统主要参数对多波形识别效果的影响.在最优系统参数条件下,对3种波形和9个波识别的归一化均方根误差分别达到了0.11和0.22.%Although the artificial neural network plays an important role in the fields of pattern recognition and artificial intelligence,it still has some disadvantages such as the complex structure, the time-consuming training,the requirement of a mass of hardware resource.A simple chaotic reservoir computing system based on Mackey-Glass nonlinear equations is designed,where a large number of chaos states at different time are regarded as virtual nodes to simplify the system structure. The system is employed to achieve the recognition of three kinds of waveforms and nine waves with three frequencies respectively and the impact of system parameters on the recognition performance of waveforms is analyzed.Under the circumstances of optimized system parameters,the normalized root-mean-square errors for the recognition of the three kinds of waveforms and the nine wave are 0.1 1 and 0.22 respectively.