论文部分内容阅读
构造出一个具有较多生物学背景的神经网络模型。所使用的神经元模型除了具有一般模型都有的内部电位、外部刺激、输出和阈值等数值之外,还具有局部小电位叠加、阈值动态变化等性质。利用Hebb型学习规则,在神经元间建立有向联接记忆符号串信息,网络的信息容量很大。基于神经元的性质和网络的联接情况,网络的演化具有丰富的动力学现象。网络记忆的符号串信息通常存在于网络的空间不稳定模式上,使神经元间联接权合理地动态变化,常常可以把不稳定模式稳定下来,使网络准确、无歧义地表示出期望联想的信息,实现有效的联想记忆。
Construct a neural network model with more biological background. In addition to the general model has the internal potential, external stimuli, output and threshold values, the neuron model used also has the property of local small potential superposition and threshold dynamic change. Using Hebb-type learning rules, a series of directed connection memory symbol information is established between neurons, and the network has a large information capacity. Based on the nature of neurons and the connectivity of the network, the evolution of the network has a wealth of dynamical phenomena. The symbolic string information of network memory usually exists in the space unsteady mode of the network, so that the right of connection between neurons reasonably dynamically changes. It is often possible to stabilize the unstable mode so that the network can accurately and unambiguously represent the expected association information , To achieve effective associative memory.