A simplified computational memory model from information processing.

Lanhua Zhang,Xiaoqian Ding,Yan Wang,Yiyuan Tang,Dongsheng Zhang,Yuqin Deng,Baoliang Sun

doi:10.1038/srep37470

Abstract

This paper is intended to propose a computational model for memory from the view of information processing. The model, called simplified memory information retrieval network (SMIRN), is a bi-modular hierarchical functional memory network by abstracting memory function and simulating memory information processing. At first meta-memory is defined to express the neuron or brain cortices based on the biology and graph theories, and we develop an intra-modular network with the modeling algorithm by mapping the node and edge, and then the bi-modular network is delineated with intra-modular and inter-modular. At last a polynomial retrieval algorithm is introduced. In this paper we simulate the memory phenomena and functions of memorization and strengthening by information processing algorithms. The theoretical analysis and the simulation results show that the model is in accordance with the memory phenomena from information processing view.

Highlights

This section introduces the theoretical background and key realization technology, which can help us understand the implementation and analysis of the SMIRN.Representation and algorithm.Many designers of memory model, like Tsukada[11] and so on, think that the contribution of modeling for memory lies in the representation and algorithm
The differences are obvious; first we introduce the forgetting in SMIRN which makes the memory more reasonable and perfect, and with forgetting the memory becomes a complete cognitive activity; second is the data structure in retrieval algorithm and the process in Fig. 3 is more reasonable, i.e., all the nodes on the path directly connect to root is not so reasonable and a little difficult whether from the complements or from memory phenomena
We propose a simplified computational memory model with bi-modular hierarchical structure from the information processing view based on complex networks and computer algorithms including memory formation and retrieval

Summary

Introduction

Many designers of memory model, like Tsukada[11] and so on, think that the contribution of modeling for memory lies in the representation and algorithm. Representation in computer is the data structure, such as tree[17] or vector[25]. In SMIRN, meta-memory is the basic unit, which expresses the information capability; it represents a flexible measurement, not a biological sense. The abstract data structure of MFset[26] is similar in conceptual meaning, so we select set as meta-memory data structure and tree as the network structure. Set itself can store the information of meta-memory, and the functions of memory can be expressed by the results and performance of the information retrieval from set, so we need to introduce the retrieval algorithm of set

Objectives

Methods

Discussion

Conclusion