A local distribution net for data clustering

Authors:
Qiubao Ouyang;Furao Shen;Jinxi Zhao
Affiliations:
National Key Laboratory for Novel Software Technology, Nanjing University, China;National Key Laboratory for Novel Software Technology, Nanjing University, China;National Key Laboratory for Novel Software Technology, Nanjing University, China
Venue:
PRICAI'12 Proceedings of the 12th Pacific Rim international conference on Trends in Artificial Intelligence
Year:
2012

Citing 9
Cited 0

The ART of Adaptive Pattern Recognition by a Self-Organizing Neural Network

Computer
Dimension reduction by local principal component analysis

Neural Computation
Mixtures of probabilistic principal component analyzers

Neural Computation
Unsupervised Learning of Finite Mixture Models

IEEE Transactions on Pattern Analysis and Machine Intelligence
Advanced Methods in Neural Computing

Advanced Methods in Neural Computing
An incremental network for on-line unsupervised classification and topology learning

Neural Networks
An enhanced self-organizing incremental neural network for online unsupervised learning

Neural Networks
Adaptive nonlinear manifolds and their applications to pattern recognition

Information Sciences: an International Journal
Local PCA algorithms

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, a local distribution neural network is proposed for data clustering. This competing network is designed for non-stationary and evolving environment. It represents data by means of neurons (ellipsoids) arranged on a topology map. The local distribution is stored in ellipsoids, while the global topology information is preserving in the relationship between adjacent ellipsoids. With a self-adapting threshold strategy and iteratively learning for information of local distribution, the algorithm is operated in an incremental and on-line way. During implementation, The adopted metric is an improved Mahalanobis distance which considers the local distribution and implies the anisotropy on different vector basis. Hence it can be interpreted as an incremental version of Gaussian mixture model. Experiments both on artificial data and real-world data are carried out to show the performance of the proposed method.