Stochastic Weight Update in Neural Networks: Theoretical study of stochastic neural networks learning

Ko cák, Juraj, Jak a, Rudolf, Sincák, Peter

ISBN 10: 3659231029 / ISBN 13: 9783659231025

Published by LAP LAMBERT Academic Publishing, 2012

About this Item

LIKE NEW. SHIPS FROM MULTIPLE LOCATIONS.

Seller Inventory # ERICA75836592310296

Bibliographic Details

• Title: Stochastic Weight Update in Neural Networks: Theoretical study of stochastic neural networks learning
• Publisher: LAP LAMBERT Academic Publishing
• Publication Date: 2012
• Language: English
• ISBN 10: 3659231029
• ISBN 13: 9783659231025
• Binding: Paperback
• Condition: Like New
• Book Type: book

About this title

Synopsis

This book is focused on the modification of the Backpropagation Through Time algorithm and its implementation on the Recurrent Neural Networks. Our work is inspired and motivated by the results of the Salvetti and Wilamowski experiment focused on the introduction of stochasticity into Backpropagation algorithm on experiments with the XOR problem. The stochasticity can be embedded into different parts of the BP algorithm. We introduced and implemented different types of BP algorithm modifications, which gradually add more stochasticity to the BP algorithm. The goal of this book is to prove, that this stochastic modification is able to learn efficiently and the results are comparable to classical implementation. This stochasticity also brings a simpler implementation of the algorithm, than the classical one, which is especially useful on the Recurrent Neural Networks.

About the Author

Juraj Košcak works as Senior Developer commercially.He studied his Master in 2007 and finished Doctorate part-time in 2012from Technical University in Košice,Department of Cybernetic and Artificial Intelligence, Slovakia.