Learning Exemplar-Represented Manifolds in Latent Space for Classification
Abstract
Intrinsic manifold structure of a data collection is valuable information for classification task. By considering the manifold structure in the data set for classification and with the sparse coding framework, we propose an algorithm to: (1) find exemplars from each class to represent the class-specific manifold structure, in which way the object-space dimensionality is reduced; (2) simultaneously learn a latent feature space to make the mapped data more discriminative according to the class-specific manifold measurement. We call the proposed algorithm Exemplar-represented Manifold in Latent Space for Classification (EMLSC). We also present the nonlinear extension of EMLSC based on kernel tricks to deal with highly nonlinear situations. Experiments on synthetic and real-world datasets demonstrate the merit of the proposed method.
Cite
Text
Kong and Wang. "Learning Exemplar-Represented Manifolds in Latent Space for Classification." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2013. doi:10.1007/978-3-642-40994-3_16Markdown
[Kong and Wang. "Learning Exemplar-Represented Manifolds in Latent Space for Classification." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2013.](https://mlanthology.org/ecmlpkdd/2013/kong2013ecmlpkdd-learning/) doi:10.1007/978-3-642-40994-3_16BibTeX
@inproceedings{kong2013ecmlpkdd-learning,
title = {{Learning Exemplar-Represented Manifolds in Latent Space for Classification}},
author = {Kong, Shu and Wang, Donghui},
booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
year = {2013},
pages = {240-255},
doi = {10.1007/978-3-642-40994-3_16},
url = {https://mlanthology.org/ecmlpkdd/2013/kong2013ecmlpkdd-learning/}
}