By Topic

A novel approach to optimize the objective function based on VC dimension and structural risk minimization

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

3 Author(s)
Qiu Xintao ; Sch. of Autom., Univ. of Sci. & Technol. Beijing, Beijing, China ; Fu Dongmei ; Yang Tao

So far, lots of algorithms and learning methods are taking the empirical risk for the optimization goal. In this paper, we propose a new way to optimize the objective function based on VC dimension and structural risk minimization. The optimized function F is firstly defined by us, and some effective design forms of it will also be given. Then we fulfill a useful criterion-the balance minimum optimization principle. This principle considers not only the empirical risk, but also considers the VC dimension of learning machine. Balancing the two factors can avoid the underfitting problem and overfitting problem. Experimental results show that the method we proposed is effective to improve the property of algorithm efficiency, convergence and generalization of the learning machine. Also, the proposed principle for optimization is a new criterion, which is not a practical method for a particular problem of improvement. Therefore, this method is suitable to be applied in many practical situations, which may bring a good generalization performance and efficiency in some learning problems.

Published in:

Control Conference (CCC), 2011 30th Chinese

Date of Conference:

22-24 July 2011