| 杨南海,桑媛媛,赫然,王秀坤.基于非负稀疏表示的标签繁殖算法[J].,2012,(2):264-271 |
| 基于非负稀疏表示的标签繁殖算法 |
| Label propagation algorithm based on nonnegative sparse representation |
| |
| DOI：10.7511/dllgxb201202018 |
| 中文关键词: 非负稀疏表示 半监督学习 稀疏概率图 聚类关系 标签繁殖 |
| 英文关键词: nonnegative sparse representation semi supervised learning sparse probability graph clustering relationship label propagation |
| 基金项目:高等学校博士学科点专项科研基金资助项目(20100041120009);国家自然科学基金资助项目(60873054);大连理工大学引进人才启动经费资助项目. |
|
| 摘要点击次数: 1635 |
| 全文下载次数: 957 |
| 中文摘要: |
| 提出了一种基于非负稀疏表示(nonnegative sparse representation,NSR)的半监督学习标签传播算法．该算法首先构造一个稀疏概率图(sparse probability graph,SPG),其权重由非负稀疏表示算法计算的非负系数组成,自然地反映了各样本之间的聚类关系,避免了传统半监督学习算法中的邻居选择和参数设置过程；然后通过对未标记样本的标签进行迭代繁殖至收敛而获得所有样本的标签．在人脸识别、物体识别、UCI机器学习和TDT文本数据集上的实验结果表明采用非负稀疏表示的标签传播算法比典型的标签繁殖算法具有更好的分类准确率． |
| 英文摘要: |
| A novel label propagation algorithm for semi-supervised learning based on nonnegative sparse representation(NSR) is presented. Firstly, this algorithm derives a sparse probability graph (SPG) from nonnegative weight coefficients which are computed by nonnegative sparse representation algorithm. The weights of SPG naturally reveal the clustering relationship of labeled samples and unlabeled samples, meanwhile avoid the adjacency selection and parameter setting process in traditional semi supervised learning algorithm. Then, the labels of unlabeled samples are propagated until convergence to obtain all the labels of samples. Extensive experimental results on face recognition, object recognition, UCI machine learning and TDT text datasets show that label propagation algorithm based on NSR achieves the lower error rate as compared with the standard label propagation algorithm. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|