By Topic

An Improved Adaptive Boosting Algorithm for Neural Network Ensemble Based on Multi-dimensional Cloud Model

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Zhigang Gan ; Sch. of Comput. Sci. & Eng., South China Univ. of Technol., Guangzhou, China ; Xiao, Nanfeng

AdaBoosting is widely used in neural network ensemble as a variety of boosting algorithm. However, with the learning pattern of focusing on hard sample, AdaBooting makes neural network fall into degradation easily. Additionally, in neural network ensemble the weight of individual neural network only takes into account the misclassifying rate. In fact, due to the characteristic of neural network's tendency of learning hard samples in Adaboosting algorithm, the predictive accuracy of individual neural network's training for a particular sample space is much higher than for the whole sample space. This paper proposes an improved AdaBoosting algorithm called Cloud-AdaBoosting. In this method, it is introduced the concept of cloud model and applied the technique of filling training set with similar samples generated by cloud generator to overcome the degradation phenomena resulted from the tendency of learning hard samples. Through calculating the certainty degree of each test sample relative to neural network, it is adjusted dynamically the weight of whole neural network output by individual neural network which can reflect the prediction ability of a part of samples in neural network ensemble, and then enhance the whole prediction performance of the ensemble. The experiment results show that the proposed algorithm is efficient to increase the prediction accuracy of neural network ensemble.

Published in:

Knowledge Acquisition and Modeling, 2009. KAM '09. Second International Symposium on  (Volume:2 )

Date of Conference:

Nov. 30 2009-Dec. 1 2009