Semi-Supervised Regression using Cluster Ensemble and Low-Rank Co-Association Matrix

What is it about?

In this paper, we consider a semi-supervised regression problem in the transductive learning setting. Due to the luck of knowledge about the data structure and the presence of random noise, the considered data model is uncertain. We propose a method which combines graph Laplacian regularization and cluster ensemble methodologies. The co-association matrix of the ensemble is calculated on both labeled and unlabeled data; this matrix is used as a similarity matrix in the regularization framework to derive the predicted outputs. We use the low-rank decomposition of the co-association matrix to significantly speedup calculations and reduce memory. Numerical experiments include: 1) an artificial example with two clusters and random noise, 2) 10-dimensional real forest fires dataset example.

Featured Image

Photo by DDP on Unsplash

Photo by DDP on Unsplash

Why is it important?

The task of semi-supervised learning is important because in many real-life problems only a small part of available data can be labeled due to the large cost of target feature registration. For example, manual annotation of digital images is rather time-consuming. Therefore labels can be attributed to only a small part of pixels. To improve prediction accuracy, it is necessary to use information contained in both labeled and unlabeled data. An important application is hyperspectral image semi-supervised classification.

Perspectives