A Cognitive Model Based Framework and Multi-layer Storage Architecture for Associative Memory

Abstract

Memory is the foundation of intelligence. KID model covers multiple human cognitive processes such as learning and memory. This paper refines its memory process, especially focusing on associative memory of long-term memory. An associative memory framework with novel neural network storage architectures is presented to simulate human-like associative memory ability for machine intelligence. The presented framework involves an associative memory repository and two abstract functions, Assimilation() for knowledge encoding and storage, and Instantiation() for knowledge recall and application. The proposed novel storage architecture has two storage structures which both includes three kinds of layers: input layer, competitive layer and associative memory layer. Its design integrates multiple associative memory related neuroscience theories. It is characterized by chaotic feature, self-organization, self-adjustment, self-growth and associative recall. With the encapsulation of presented associative memory framework and novel storage architecture, the KID model can incorporate associative memory and be applied to various fields like intelligent information and knowledge management systems, personized products development and robotic intelligence.