ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

An Endowed Takagi-Sugeno-type Fuzzy Model for Classification Problems?

Journal: International Journal of Computer Science and Mobile Computing - IJCSMC (Vol.3, No. 11)

Publication Date:

Authors : ; ;

Page : 201-206

Keywords : Batch support vector machine; high generalization ability; new incremental learning approach; one-pass clustering algorithm; online training samples;

Source : Downloadexternal Find it from : Google Scholarexternal


Takagi-Sugeno-type fuzzy classification model is improved by proposing a new incremental learning approach. Margin selective gradient descent learning and incremental support vector machine helps the proposed fuzzy model to learn with high generalization ability. Training samples are provided incrementally one after another instead of in a single batch. This improved fuzzy model is obtained from an empty set and clustering algorithm is used for finding initial fuzzy sets and number of rules in the antecedent part. The generalization ability of the fuzzy model can be improved by proposing an online incremental linear support vector machine to tune the rule consequent parameters. Online training problem can be solved by incremental support vector machine, where only one new training sample is provided at a time. Margin selective gradient descent algorithm is used to learn antecedent parameter and avoids overtraining.

Last modified: 2014-11-18 03:07:03