|
This article is cited in 1 scientific paper (total in 1 paper)
Deep learning neural network structure optimization
M. S. Potanina, K. O. Vaysera, V. A. Zholobova, V. V. Strijovba a Moscow Institute of Physics and Technology, 9 Institutskiy Per., Dolgoprudny, Moscow Region 141700, Russian Federation
b A. A. Dorodnicyn Computing Center, Federal Research Center “Computer Science and Control” of the Russian Academy of Sciences, 40 Vavilov Str., Moscow 119333, Russian Federation
Abstract:
The paper investigates the optimal model structure selection problem. The model is a superposition of generalized linear models. Its elements are linear regression, logistic regression, principal components analysis, autoencoder and neural network. The model structure refers to values of structural parameters that determine the form of final superposition. This paper analyzes the model structure selection method and investigates dependence of accuracy, complexity and stability of the model on it. The paper proposes an algorithm for selection of the neural network optimal structure. The proposed method was tested on real and synthetic data. The experiment resulted in significant structural complexity reduction of the model while maintaining accuracy of approximation.
Keywords:
model selection, linear models, autoencoders, neural networks, structure, genetic alghorithm.
Received: 02.12.2019
Citation:
M. S. Potanin, K. O. Vayser, V. A. Zholobov, V. V. Strijov, “Deep learning neural network structure optimization”, Inform. Primen., 14:4 (2020), 55–62
Linking options:
https://www.mathnet.ru/eng/ia697 https://www.mathnet.ru/eng/ia/v14/i4/p55
|
|