摘要
在对联想记忆神经网络的研究中,为提高现有联想记忆网络的存储能力以及相似模式和多值模式的联想成功率,提出了一种新的联想记忆网络。样本模式信息存储在动态权值矩阵中,网络根据不同的输入模式可自适应地调节当前权值矩阵。与传统联想网络相比,输入模式的信息不仅给出了联想记忆的初值,且在联想记忆过程中起到启发式搜索的作用,使网络的存储能力和联想成功率得到较好的改善。尤其可以有效地实现相似模式以及多值模式的联想记忆功能。仿真结果验证了方法的有效性。
Combining with the theory of cognitive process, a novel method for associative memory is proposed to enlarge the memory capability and enhance the associative success rate. Dynamic connection weights are designed to store the information of the sample patterns. The weights can be chosen dynamically according to the current input pattern. The information of input pattern not only provides the initial values for the associative memory, but also plays heuristic searching role in the associative process, which can enhance the memory capability and the associative suc- cess rate. Furthermore, the associative memory for the similar patterns and multi -values patterns can also be com- pleted by the method. The simulation results prove the validity of the algorithm.
出处
《计算机仿真》
CSCD
北大核心
2010年第4期176-179,共4页
Computer Simulation
基金
天津市高等学校科技发展基金(20060613)
国家自然科学基金(60808020)
关键词
联想记忆
多值模式
相似模式
存储容量
Associative memory
Multi - values pattern
Similar pattern
Memory capability