Comparative analysis of gradient minimization methods in the task of multilayer perceptron learning

Abstract: The article presents the comparative analysis of gradient minimization methods of target error function in the task of neural network training. The description and characteristics of these methods are given. The results of the analysis and numerical experiments demonstrate that the conjugate gradient method is most efficient for neural network training.

Keywords: neural network, gradient methods

For citation:
Skorohodov A. V., Tungusova A. V. Comparative analysis of gradient minimization methods in the task of multilayer perceptron learning. Doklady Tomskogo gosudarstvennogo universiteta sistem upravleniya i radioelektroniki, 2011, no. 2(24), – p. 3. pp. 98–102.