Proceedings

A New Simplex Sparse Learning Model to Measure Data Similarity for Clustering

Jin Huang, Feiping Nie, Heng Huang

Download BibTeX Cite

Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence. Special Track on Machine Learning. Pages 3569-3574.

Abstract

The Laplacian matrix of a graph can be used in many areas of mathematical research and has a physical interpretation in various theories. However, there are a few open issues in the Laplacian graph construction: (i) Selecting the appropriate scale of analysis, (ii) Selecting the appropriate number of neighbors, (iii) Handling multiscale data, and, (iv) Dealing with noise and outliers. In this paper, we propose that the affinity between pairs of samples could be computed using sparse representation with proper constraints. This parameter free setting automatically produces the Laplacian graph, leads to significant reduction in computation cost and robustness to the outliers and noise. We further provide an efficient algorithm to solve the difficult optimization problem based on improvement of existing algorithms. To demonstrate our motivation, we conduct spectral clustering experiments with benchmark methods. Empirical experiments on 9 data sets demonstrate the effectiveness of our method.

Jin Huang, Feiping Nie, and Heng Huang. 2015. A New Simplex Sparse Learning Model to Measure Data Similarity for Clustering. In Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-15. IJCAI, 3569–3574. https://ijcai.org/proceedings/2015/502
Welcome Future Conferences AI Journal Division IJCAI Proceedings About IJCAI IJCAI Awards