双向联想记忆神经网络通常用于描述借助于双向,即前后两个方向,储存或记忆一对类似模式的能力,这方面的研究对于模式识别以及自动控制工程中的应用是至关重要的,而网络系统的同步是近年来国内外研究的热点。本文采用摄动一增量法解析定量地研究了具有四个神经元和两个离散时滞的BAM神经网络模型的同步周期解,不仅提出了所研究的时滞耦合网络系统同步周期解的充分必要条件,而且还给出了由时滞诱发的完全同步周期解的近似解析形式。最后通过数值模拟进一步说明本文所采用的方法是有效的,所得的结果是正确的。本文的结果在设计人工神经网络方面有着潜在的应用。
The bi-directional associative memory neural networks have very important applications in storing paired patterns or memories and possess the ability of searching the desired patterns via both direc- tions, forward and backward, such as the pattern recognition and automatic control Engineering. However, the network synchronization issues have attached worldwide attention in recent years. A bi-directional associative memory neural network with four neurons and two discrete delays was considered quantitatively by using the perturbation-incremental scheme. The necessary and sufficient conditions of synchronized periodic solution were provided from a Hopf bifurcation and the synchronized periodic solution was given analytically. The validity and correction of the obtained analytical results are shown by their consistency with numerical simulation. The results have some potential applications, such as designing neural networks.