Solomatin V F
St. Petersburg State University, Russia.
Biofizika. 1999 Jan-Feb;44(1):123-7.
Three-layer models with neural plasticity are considered. The weights of connections have both positive and negative values, which are chosen randomly with equal probability. The connections are symmetric relative to the layer of associative neurons. The models describe processes analogous to holographic ones, and the basic properties of the models correspond to the properties of brain memory. The random choice of weights providing the noncorrelatedness of separate recordings gives rise to disturbances. It was shown that in passing from a full-connected model to non-full-connected and semi-full-connected models, the disturbances increase and limitations arise. The degree of the increase in disturbances depends upon the type of the model, the methods of recording and methods of the use of the models, and the mechanisms of the origination of disturbances.
考虑具有神经可塑性的三层模型。连接权重具有正值和负值,它们以相等的概率随机选择。连接相对于联想神经元层是对称的。这些模型描述了类似于全息过程的过程,并且模型的基本属性与大脑记忆的属性相对应。权重的随机选择提供了单独记录的不相关性,从而产生了干扰。结果表明,从全连接模型转变为非全连接和半全连接模型时,干扰增加且出现了限制。干扰增加的程度取决于模型的类型、记录方法和模型的使用方法以及干扰产生的机制。