摘要
为减轻用户疲劳并将交互式遗传算法应用于复杂的优化问题中,提出一种基于半监督支持向量机的交互式遗传算法。根据标记样本和未标记样本几何特性派生出数据依赖的核函数,以此构建半监督支持向量机,再以自训练方法进行高可信未标记样本的批量选择,实现用户评价代理模型的高泛化性能。将该方法应用于基于内容的图像检索系统,结果表明其能有效加快进化收敛的速度,提高优化成功率。
In order to alleviate user fatigue and apply the interactive Genetic Algorithm(GA) into complicated optimization problems, this paper presents interactive GA based on Semi-supervised Support Vector Machine(S3VM), which is used to establish the surrogate model. According to the geometry of the underlying marginal distribution from both labeled data and unlabeled data, it derives a data-dependent kernel in order to establish S3VM. Self-training method is employed for batch selecting the high reliable unlabeled samples. The method is applied to relevance feedback image retrieval, and experimental results show it is effective to accelerate the evolution of the convergence and increases the optimization success ratio.
出处
《计算机工程》
CAS
CSCD
2012年第21期182-184,188,共4页
Computer Engineering
基金
国家自然科学基金资助项目(61170038)
关键词
交互式遗传算法
半监督学习
支持向量机
核函数
代理模型
用户疲劳
Interactive Genetic Algorithm(IGA)
semi-supervised learning
Support Vector Machine(SVM)
kernel function
surrogate model
user fatigue