Duch W and Korczak J (1999), Optimization and global minimization methods suitable for neural networks, Neural Computing Surveys. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Meta-Learning Evolutionary Artificial Neural Networks - Abraham (2003) (Correct)
.... the first attempt of the evolution of learning rules [40] Chalmers [23] defined the form of learning rules as a linear function of four local variables and their six pair wise products [11] 33] Global optimization of neural network has been widely addressed using several other techniques [22] [28] [34] 64] 71] 72] 73] 74] 86] Sexton et al. [72] used simulated annealing algorithm for optimization of learning. For optimization of the neural network learning, in many cases a pre defined architecture was used and in a few cases architectures were evolved together. No work has been ....
Duch W and Korczak J (1999), Optimization and global minimization methods suitable for neural networks, Neural Computing Surveys.
Searching for optimal MLP - Duch, Grabczewski (1999) Self-citation (Duch) (Correct)
....in which gradients are non zero, shrinks rapidly to zero when sigmoids are changed into step functions. In effect linear discrimination analysis may sometimes obtain better results than non linear neural techniques [1] To avoid such drawbacks one may use either global minimization techniques [2], which from computational point of view are rather expensive, use better initialization methods [3] 1] or try to average over an ensemble of neural models [4] 2] In this paper a radically different approach is proposed. Minimization and search methods [5] share the same goal of optimizing ....
.... better results than non linear neural techniques [1] To avoid such drawbacks one may use either global minimization techniques [2] which from computational point of view are rather expensive, use better initialization methods [3] 1] or try to average over an ensemble of neural models [4] [2]. In this paper a radically different approach is proposed. Minimization and search methods [5] share the same goal of optimizing some cost functions. Quantization of network parameters (weights and biases) allows to replace minimization by search. Increasing step by step the resolution of ....
W. Duch, J. Korczak, Optimization and global minimization methods suitable for neural networks. Neural Computing Surveys (submitted).
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC