Nonnegative Sparse and KNN graph for semi-supervised learning

DOI

10.2991/ameii-16.2016.223How to use a DOI?

Keywords

Sparse graph, KNN graph, NSKNN-graph, semi-supervised learning

Abstract

For the graph-based semi-supervised learning, the performance of a classifier is very sensitive to the structure of the graph. So constructing a good graph to represent data, a proper structure for the graph is quite critical. This paper proposes a novel model to construct the graph structure for semi-supervised learning. In this new structure, the weights of edges in the graph are obtained by the linear combination of a Nonnegative Sparse graph and K Nearest Neighbour graph (NSKNN-graph). The NSKNN-graph can capture both the global structure (by global sparse graph) and the local structure (by the KNN graph). We demonstrate the effectiveness of NSKNN-graph on the UCI dataset. Experiments show that the NSKNN-graph has advantages over graphs constructed by conventional methods.

Copyright

© 2016, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Yunbin Zhang
AU  - Chunmei Zhang
AU  - Qianqi Zhou
PY  - 2016/04
DA  - 2016/04
TI  - Nonnegative Sparse and KNN graph for semi-supervised learning
BT  - Proceedings of the 2nd International Conference on Advances in Mechanical Engineering and Industrial Informatics (AMEII 2016)
PB  - Atlantis Press
SP  - 1178
EP  - 1182
SN  - 2352-5401
UR  - https://doi.org/10.2991/ameii-16.2016.223
DO  - 10.2991/ameii-16.2016.223
ID  - Zhang2016/04
ER  -