Comparative analysis of the results of training a neural network with calculated weights and with random generation of the weights

DOI10.1134/S0005117920070048zbMATH Open1497.68421MaRDI QIDQ828095FDOQ828095

Authors: P. Sh. Geidarov 

Publication date: 14 January 2021

Published in: Automation and Remote Control (Search for Journal in Brave)

Recommendations

• The use of random weights for the training of multilayer networks of neurons with Heaviside characteristics
• Algorithm of choosing initialized weights for training complex-valued neural networks
• The sample complexity of pattern classification with neural networks: the size of the weights is more important than the size of the network
• Structural, Syntactic, and Statistical Pattern Recognition
• Multitasking application of neural networks implementing metric methods of recognition

zbMATH Keywords

neural networksnearest neighbor methodbackpropagation algorithmsmetric recognition methodsrandom weight initialization

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Machine vision and scene understanding (68T45)

Cites Work

• Title not available (Why is that?)
• Multitasking application of neural networks implementing metric methods of recognition

Cited In (1)

• Multitasking application of neural networks implementing metric methods of recognition

This page was built for publication: Comparative analysis of the results of training a neural network with calculated weights and with random generation of the weights

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q828095)