一种基于半监督多任务学习的特征选择模型

A Feature Selection Framework Based on Semi-supervised Multi-task Learning

为了综合利用流形学习、多任务学习和正则化约束的优势,提出一种基于全局和局部约束的半监督多任务特征选择(semi-supervised multi-task feature selection,SMFS)模型,在多个任务间共享学习的基础上,构建SMFS模型.该模型采用l2,1范数约束选择最具判别性的特征,避免噪声的干扰,并引入局部信息约束提高特征选择的准确度.将SMFS模型应用于网页自动分类,与目前流行的几种算法进行对比,证明了该算法的有效性.

Feature selection,which aims to reduce the dimension of the data and remove the redundant feature,plays an important role in improving the performance of multimedia processing.In this paper,a semi-supervised multi-task feature selection algorithm built on sharing information between multiple learning tasks has been proposed.In order to select the most discriminative features,and avoid the noise interference,we have also constructed a semi-supervised multi-task feature selection model with l2,1-norm and local information constraint.In order to verify the effectiveness of our algorithm,we apply the algorithm to the web page classification application and compare it with several state-of-the-art algorithms.Results show that the proposed algorithm is effective.