Geodesic
On the optimization of the constructive method of training neural networks
Vestnik KRAUNC. Fiziko-matematičeskie nauki, no. 3 (2018), pp. 184-189 Cet article a éte moissonné depuis la source Math-Net.Ru

Voir la notice de l'article

The article suggests a constructive method for training neural networks in which neurons added just before the current epoch of training assume the main training load on the new class to ensure the stability of the network in relation to learning on new data classes. The results of computational experiments are presented.
Keywords: neural networks, machine learning, constructive training methods. 
@article{VKAM_2018_3_a21,
     author = {M. A. Kazakov},
     title = {On the optimization of the constructive method of training neural networks},
     journal = {Vestnik KRAUNC. Fiziko-matemati\v{c}eskie nauki},
     pages = {184--189},
     year = {2018},
     number = {3},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/VKAM_2018_3_a21/}
}
TY  - JOUR
AU  - M. A. Kazakov
TI  - On the optimization of the constructive method of training neural networks
JO  - Vestnik KRAUNC. Fiziko-matematičeskie nauki
PY  - 2018
SP  - 184
EP  - 189
IS  - 3
UR  - http://geodesic.mathdoc.fr/item/VKAM_2018_3_a21/
LA  - ru
ID  - VKAM_2018_3_a21
ER  - 
%0 Journal Article
%A M. A. Kazakov
%T On the optimization of the constructive method of training neural networks
%J Vestnik KRAUNC. Fiziko-matematičeskie nauki
%D 2018
%P 184-189
%N 3
%U http://geodesic.mathdoc.fr/item/VKAM_2018_3_a21/
%G ru
%F VKAM_2018_3_a21
M. A. Kazakov. On the optimization of the constructive method of training neural networks. Vestnik KRAUNC. Fiziko-matematičeskie nauki, no. 3 (2018), pp. 184-189. http://geodesic.mathdoc.fr/item/VKAM_2018_3_a21/

[1] SHibzuhov Z. M., Konstruktivnye metody obucheniya $\Sigma\Pi$ - nejronnyh setej, Nauka, Moskva, 2006, 159 pp. | MR

[2] Timofeev A.V., Kosovskaya T.M., “Nejrosetevye metody logicheskogo opisaniya i raspoznavaniya slozhnyh obrazov”, Trudy SPIIRAN, 27 (2013), 144-155

[3] LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton, “Deep learning”, Nature, 521:7553 (2015), 436-444

[4] Sutskever I., Martens J., Dahl G., Hinton G., “On the importance of initialization and momentum in deep learning”, J. of Machine Learning Research, 28 (2013), 1139-1147