Abstract:
Aiming at the advantages of rapidly converging and light load inside for Sequential Minimal Optimization(SMO), this paper transplants it into 1-Fuzzy Support Vector Machine(1-FSVM). In order to enhance the training speed, 1-FSVM algorithm uses hierarchical Binary tree structure to cluster step by step, takes different weighting in every level for different input vector to correctly express the classification effect. Application result in light recognition handwritten numeral sets and license plate location show that 1-FSVM algorithm has a high detection rate and speed.
Key words:
1-Fuzzy Support Vector Machine(1-FSVM),
Sequential Minimal Optimization(SMO),
hierarchical
摘要: 针对序贯最小优化(SMO)训练算法具有计算速度快、无内负荷的特点,将其移植到模糊一类支持向量机(1-FSVM)中。1-FSVM算法融入层次型偏二叉树结构进行逐步聚类以加快训练速度,并对每个输入向量赋予不同权值以达到准确的分类效果。应用于光识别手写数字集和车牌定位的结果表明,1-FSVM算法具有较高的检测率与较快的检测速度。
关键词:
模糊一类支持向量机,
序贯最小优化,
层次型
CLC Number:
ZUO Ping-Beng, SUN Bin, GU Hong, JI Dong-Lian. Hierarchical 1-FSVM Algorithm Based on SMO[J]. Computer Engineering, 2010, 36(19): 188-189,192.
左萍平, 孙赟, 顾弘, 齐冬莲. 基于SMO的层次型1-FSVM算法[J]. 计算机工程, 2010, 36(19): 188-189,192.