摘要
基于传统感知器在分类中的优势,结合新型的电路元件——忆阻器作突触,提出新型的忆阻感知器。推导了忆导变化与权值更新的关系,构建了单层忆阻感知器和多层忆阻感知器模型,提出了忆阻突触的权值更新规则。通过Matlab仿真,用单层忆阻感知器实现了线性可分逻辑"与"和逻辑"或"分类问题,用多层忆阻感知器实现了线性不可分的逻辑"异或"和"同或"分类问题,从而证实了该方案的有效性。
A new memristive perceptron was proposed based on the advantages of perceptron in logical classification combining with memristors, new circuit elements. The relationship between memristive conductance and synapse weight updating was derived. The modeling of single-layer memristive perceptron and multi-layer memristive perceptron were built, and weight updating rule of memristive synapse was proposed. Matlab simulation confirmed that the memristive perceptron can realize the logical classification. That is to say, a single-layer memristive perceptron can realize the classification of linear separable logical AND and OR, and the linear inseparable logical XOR and NXOR can be implemented with muhilayer memristive perceptron. Memristive perceptron could be used for solving complex artificial intelligence problems.
出处
《重庆理工大学学报（自然科学）》
CAS
2013年第1期71-75,94,共6页
Journal of Chongqing University of Technology：Natural Science
基金
国家自然科学基金资助项目(60972155
61101233
60974020)
中央高校基本科研业务费专项(XD-JK2012A007
XDJK2010C023)
重庆市高等学校青年骨干教师资助计划(2011-65)
重庆市高等学校优秀人才支持计划(2011-65)
中国博士后科学基金资助项目(CPSF20100470116)
重庆市高等教育教学改革研究重点项目(09-2-011)
关键词
忆阻器
感知器
逻辑分类
memristor
perceptron
logical classification
