A hierarchically organized memory model with temporal population coding

Memory is a critical process in the brain to many cognitive behaviors. It is a complex process operating across different brain regions. However, the organizing principles of memory systems remain unclear. Emerging experiment results show that memories are represented by population of neurons and organized in a categorical and hierarchical manner. In this work, we describe a hierarchically organized memory (HOM) model using spiking neurons, in which temporal population codes are considered as the neural representation of information and spike-timing-based learning methods are employed to train the network. The results have demonstrated that memory coding units are formed into neural cliques, and information are stored in the form of associative memory within gamma cycles. Moreover, temporally sepa-rated patterns can be linked and compressed via enhanced connections among neural groups forming episodic memory. Our model provides a computational interpretation of memory organization at a system level.