投影自表示无监督极限学习机

无监督极限学习机在投影过程中保持原始高维空间中的稀疏或近邻结构,样本在高维空间中存在冗余信息,原始的数据结构不一定适应于投影后的低维特征空间.为此,结合无监督极限学习机和子空间聚类的自表示学习,提出投影自表示无监督极限学习机模型.该模型是面向聚类的特征提取方法,在投影过程中学习自表示子空间结构,从而使无监督极限学习机提取的特征自适应于聚类任务.在IRIS数据集、 6个基因表达和2个医学影像高维数据集上进行实验,结果表明该模型和算法是有效的.

Projected Self-expressive Unsupervised Extreme Learning Machine

The unsupervised extreme learning machine preserves the sparsity and neighborhood in the original high-dimensional space during the projection. The sample has redundant information in the high-dimensional space, and the original data structure may not be adapted to the low-dimensional feature space after projection. This paper combines the unsupervised ELM and the self-representation learning of subspace clustering, and proposes the Projected Self-expressive Unsupervised Extreme Learning Machine (PS-ELM) method. PS-ELM is a feature extraction method for clustering, which learns the self-expressive subspace structure during the projection process, so that the features extracted by the unsupervised extreme learning machine are adaptive to the clustering task. Experiments were performed on the IRIS data set, 6 gene expressions and 2 medical imaging high-dimensional data sets. Experimental results show that the model and algorithm are effective.