Anastasiadis, A.D. and Magoulas, George D. and Vrahatis, M.N. (2006) Improved sign-based learning algorithm derived by the composite nonlinear Jacobi process. Journal of Computational and Applied Mathematics 191 (2), pp. 166-178. ISSN 0377-0427.
Download (490kB) | Preview
In this paper a globally convergent first-order training algorithm is proposed that uses sign-based information of the batch error measure in the framework of the nonlinear Jacobi process. This approach allows us to equip the recently proposed Jacobi–Rprop method with the global convergence property, i.e. convergence to a local minimizer from any initial starting point. We also propose a strategy that ensures the search direction of the globally convergent Jacobi–Rprop is a descent one. The behaviour of the algorithm is empirically investigated in eight benchmark problems. Simulation results verify that there are indeed improvements on the convergence success of the algorithm.
|Additional Information:||Copyright © 2006 Elsevier B.V.|
|Keyword(s) / Subject(s):||supervised learning, nonlinear iterative methods, nonlinear Jacobi, pattern classification, feedforward neural networks, convergence analysis, global convergence|
|School or Research Centre:||Birkbeck Schools and Research Centres > School of Business, Economics & Informatics > Computer Science and Information Systems|
|Depositing User:||Sandra Plummer|
|Date Deposited:||04 Jun 2007|
|Last Modified:||17 Apr 2013 12:33|
Archive Staff Only (login required)