@inproceedings{94e21dd88e884b9896f511ff9a6ee95f,
title = "Locality preserving projections with adaptive neighborhood size",
abstract = "Feature extraction methods are widely employed to reduce dimensionality of data and enhance the discriminative information. Among the methods, manifold learning approaches have been developed to detect the underlying manifold structure of the data based on local invariants, which are usually guaranteed by an adjacent graph of the sampled data set. The performance of the manifold learning approaches is however affected by the locality of the data, i.e. what is the neighborhood size for suitably representing the locality? In this paper, we address this issue through proposing a method to adaptively select the neighborhood size. It is applied to the manifold learning approach Locality Preserving Projections (LPP) which is a popular linear reduction algorithm. The effectiveness of the adaptive neighborhood selection method is evaluated by performing classification and clustering experiments on the real-life data sets.",
keywords = "Dimensionality reduction, Feature extraction, Locality preserving projections, Neighborhood size",
author = "Wenjun Hu and Xinmin Cheng and Yunliang Jiang and Choi, {Kup Sze} and Jungang Lou",
year = "2017",
month = jan,
day = "1",
doi = "10.1007/978-3-319-63309-1_21",
language = "English",
isbn = "9783319633084",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "223--234",
booktitle = "Intelligent Computing Theories and Application - 13th International Conference, ICIC 2017, Proceedings",
address = "Germany",
note = "13th International Conference on Intelligent Computing, ICIC 2017 ; Conference date: 07-08-2017 Through 10-08-2017",
}