符号序列在人工神经网络中的记忆、表达和回忆

MEMORY, REPRESENTATION AND RECOLLECTION OF SYMBOLIC SEQUENCE IN NEURAL NETWORK

本文旨在探讨符号序列在人工神经网络中的记忆、表达和回忆的机理。通过利用符号的音频表示、环境反馈网络、符号动力系统和激发序列等理论概念建造具体模型,着重分析了简单动力系统中的激发序列在神经网络系统中的记忆、表达和回忆的过程。本文有助于促进我们去研究层次结构更为复杂的符号序列在神经网络系统中的统一表达形式,也有助于我们去探讨人类记忆的本质。

The paper is aimed at investigating a kind of mechanism by which symbolic sequence can be memorized, represented and recollected in neural network. By applying some theoretical conceptions, such as symbolic audio-freqency representation, environmental feedback network, symbolic dynamic system and excited sequence, to construct a specific model, it concentrates on analysing how excited sequence in simple dymamic system is memorized, represented and recollected in neural network system. It will help us to study how to represent symbolic sequences with more complicated structures in a united form in neural network, and also help us to explore the nature of human memory.