巢湖学院学报 ›› 2022, Vol. 24 ›› Issue (3): 97-103.doi: 10.12152/j.issn.1672-2868.2022.03.014

• 工程与技术 • 上一篇    下一篇

联合线性重构与非负稀疏表示的标记分布学习算法

彭兴,吴其林:巢湖学院 信息工程学院   李婵,等:巢湖学院 数学与统计学院   

  1. 1. 巢湖学院 信息工程学院,安徽 巢湖 238024;2. 巢湖学院 数学与统计学院,安徽 巢湖 238024
  • 收稿日期:2022-03-23 出版日期:2022-05-25 发布日期:2022-09-07
  • 作者简介:彭兴(1989—),男,安徽合肥人,巢湖学院信息工程学院助教,主要从事大数据处理、机器学习研究。
  • 基金资助:
    安徽省重点研究与开发计划项目(项目编号:201904a05020091);安徽高校自然科学研究项目(项目编号:KJ2021A1033); 巢湖学院科学研究项目(项目编号:XLY-202110);巢湖学院学科建设质量提升工程立项建设项目(项目编号:kj20xzgx01)

Label Distribution Learning Algorithm Based on Joint Linear Reconstruction and Non-Negative Sparse Representation

PENG Xing,WU Qi-lin:School of Information Engineering, Chaohu University;LI Chan,et al:School of Mathematics and Statistics, Chaohu University   

  1. 1. School of Information Engineering, Chaohu University, Chaohu Anhui 238024; 2. School of Mathematics and Statistics, Chaohu University, Chaohu Anhui 238024
  • Received:2022-03-23 Online:2022-05-25 Published:2022-09-07

摘要: 针对标记分布学习涉及到样本的特征相关性信息及数据可能存在异常和噪声值的情况,结合样本的自我表示性质和样本与标记之间的相关性建立模型,提出联合线性重构与非负稀疏表示的标记分布学习算法(LRNSR-LDL)。首先用特征的自我表示属性,建立样本特征空间之间的线性关系,得到线性重构后的特征相似空间;然后利用特征和标记之间的相关性,通过非负稀疏矩阵分解将标记分布用特征相似空间表示,并分别用损失函数建立优化模型;最后引入l2,1-范数约束,降低离群点的不良影响,同时增加模型的泛化能力。提出算法与现有的3种标记分布学习算法在6个真实数据集上进行对比实验,并分别用5种距离和相似性指标进行评价,最终的实验结果显示提出的LRNSR-LDL算法具有一定的优势。

关键词: 标记分布学习, 线性重构, 非负稀疏表示, l2,1-范数

Abstract: Aiming at the situation that label distribution learning involves the feature correlation information of samples, and data may have abnormal and noise values, a model is established by combining the self-representation properties of samples and correlation between sample and label, and a label distribution learning algorithm (LRNSR-LDL) based on joint linear reconstruction and non-negative sparse representation is proposed. Firstly, the self-representation attribute of features is used to establish the linear relationship between the sample feature spaces, and obtain the feature similarity space after linear reconstruction; Then, using the correlation between feature and label, the label distribution is represented by the feature similarity space through non-negative sparse matrix decomposition, and optimization model is established with the loss function respectively; Finally, the l2,1-norm constraint is introduced to reduce the adverse effects of outliers and increase the model's generalization ability. The proposed algorithm is compared with three existing label distribution learning algorithms on six real data sets, and five similarity and distance indices are respectively used for evaluation. The final experimental results show that the proposed LRNSR-LDL algorithm has its advantages.

Key words: label distribution learning, linear reconstruction, non-negative sparse representation, l2,1-norm

中图分类号: 

  • TP181