Grosberg A Iu, Khrustova N V
Biofizika. 1993 Jul-Aug;38(4):726-35.
Model of neural networks system, in which networks interact by transmission and associative recognition of signals, is studied by computer simulation and qualitative approach. System behavior depends on the value of learning parameter epsilon, which determines the weight of writing in memory of each network every transmissible signal. Two different regimes are found: regime of auto-governed behavior, which depends only on initial networks characteristics, and regime of collective recognition of initial signal in form of a certain stable signals cycle. Analogy of this model and Aigen's hypercycle, the problem of creation of some new information in this model are discussed, too.
通过计算机模拟和定性方法研究了神经网络系统模型，其中网络通过信号的传输和关联识别进行交互。系统行为取决于学习参数ε的值，该参数决定了每个可传输信号在每个网络的内存写入权重。发现了两种不同的状态：自调节行为状态，仅取决于初始网络特征；以及以某种稳定信号循环形式对初始信号进行集体识别的状态。还讨论了该模型与艾根超循环的类比，以及该模型中一些新信息的创建问题。
