[WITHDRAWN] A Relevance Vector Machine Based on Gaussian Mixture Kernel

DOI

10.2991/mecae-18.2018.82How to use a DOI?

Keywords

Relevant vector machine; Gaussian mixture kernel; Gaussian mixture model.

Abstract

Relevance vector machine (RVM), a sparse Bayesian kernel method in machine learning, has been well-known for its sparsity and probabilistic predictions. Like other kernel methods, it use the kernel functions to map the input instances into higher dimensional space for problem simplicity. At present, the most widely used kernel function is Radial basis function (RBF). However, the RBF kernel does not consider the distribution information of the training samples which sometimes leads to a poor efficiency especially in semi-supervised learning where partially labeled examples are available. Therefore, in this paper, we propose a relevance vector machine base on Gaussian mixture kernel, which explores the distribution features of samples with a probabilistic density model, the Gaussian mixture model (GMM), and merge it into the training process. Applied to several datasets, the proposed method shows significantly better performance than the traditional RVM algorithm.

Copyright

© 2018, 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  - Mi Tong
AU  - Fang Liu
AU  - Quan Qi
AU  - Wangchen Qin
PY  - 2018/03
DA  - 2018/03
TI  - [WITHDRAWN] A Relevance Vector Machine Based on Gaussian Mixture Kernel
BT  - Proceedings of the 2018 International Conference on Mechanical, Electronic, Control and Automation Engineering (MECAE 2018)
PB  - Atlantis Press
SN  - 2352-5401
UR  - https://doi.org/10.2991/mecae-18.2018.82
DO  - 10.2991/mecae-18.2018.82
ID  - Tong2018/03
ER  -