摘要
针对Grira等近期提出的利用点对约束的半监督模糊聚类算法,其约束项与竞争聚类算法(CA)的目标函数之间数量级不一致,造成隶属度调整过度的问题,在重新定义目标函数的基础上提出一种改进算法,约束惩罚函数采用约束点对中两个样本新的联合表达式,使数量级与经典模糊聚类算法一致.实验结果显示,新算法的约束项与CA目标函数之间能很好地协调合作,并能通过对模糊隶属度的适度调整,实现更准确的聚类.
The semi-supervised fuzzy clustering with pairwise constraints recently proposed by Grira is analyzed.The disagreement on the magnitude order between constraint term and objective function of competitive clustering algorithm（CA） is the main cause for the overadjustment of membership values.Aiming at this problem,an improved algorithm is proposed based on a redefined objective function.Its penalty cost function introduces a new co-expression of two samples in the pairs,which has the same magnitude order as that of the typical fuzzy clustering.Experimental results show that the constraint term of the new algorithm can achieve good agreement and cooperation with the objective function of CA,and can produce more accurate clustering results by moderately enhancing or reducing the ambiguous memberships.
出处
《控制与决策》
EI
CSCD
北大核心
2010年第1期115-120,共6页
Control and Decision
基金
教育部新世纪优秀人才计划项目(NCET-06-0487)
国家自然科学基金项目(60572034
60973094)
江苏省自然科学基金项目(BK2006081)
江南大学创新团队计划项目(JNIRT0702)
关键词
半监督模糊聚类
竞争聚类算法
点对约束
惩罚代价函数
Semi-supervised fuzzy clustering
CA algorithm
Pairwise constraints
Penalty cost function
