A study of information representation in learning and memory and its application to artificial models.

在存储记忆时，感觉信息（自下而上）和知觉、意识和预测信息（自上而下）在神经网络的权值空间上相互作用。我们的研究表明，时空学习规则（STLR）和Hebb规则共存于海马CA1区的单个锥体神经元。在STLR机制中，树突上的突触重量变化是由输入神经元的局部关联（自下而上）决定的，而没有索马放电，而在Hebb机制中，索马放电是由自上而下的信息，如意识、知觉和预测（自上而下）决定的。STLR（局部）和Hebb（全局）在神经元水平上的共存可能支持这种动态过程，该过程重复自身，直到内部模型适合外部环境。这些结果有效地应用于人工模型。

In storing memory, sensory information (bottom-up) and awareness, consciousness and forecast information (top-down) interact on weight space of neural network. Our research revealed that spatio-temporal learning rule(STLR) and Hebb rule coexist in single pyramidal neurons of the hippocampal CA1 area. In STLR mechanism, synaptic weight changes on dendrite are determined by local association of input neurons (bottom-up) without soma firing whereas in Hebb mechanism the soma fires by top-down information such as awareness, consciousness and forecast (top-down). The coexistence of STLR (local) and Hebb (global) on the neuronal level may support this dynamic process that repeats itself until the internal model fits the external environment. These results were effectively applied to the artificial model.