自联想记忆神经网络能模拟人脑思维和机器智能,具有信息分布式存储和内容可寻址访问的重要特征,是人工神经网络研究的一个重要分支。介绍了开创自联想记忆神经网络研究先河的Hopfield联想记忆神经网络模型,分析了该模型的优缺点;然后在系统分析现有白联想记忆神经网络相关研究文献的基础上,从学习算法、体系结构和应用领域三个方面对自联想记忆神经网络的研究进展进行了归纳阐述;总结了自联想记忆神经网络目前存在的主要问题,并且预测了其未来的发展趋势。
As an important artificial neural network, auto-associative memory model (AM) can be employed to mimic human thinking and machine intelligence, which has massively parallel distributed configuration and content-addressable ability. In this paper, introduce in detail the Hopfield Associative Memory (HAM) neural network which has yielded a great impact on the development of auto-associative memory model, and analyze HAM' s strongpoint and drawback. Secondly, focusing on the existing relevant research literatures, present a survey of auto-associative memory models from the three aspects such as learning algorithm, network architecture and practical application; Finally,summarize the main question which auto-associative memory models are faced with at present, and forecast its future development tendency.