Detailansicht
Stochastic Weight Update in Neural Networks
Theoretical study of stochastic neural networks learning
ISBN/EAN: 9783659231025
Umbreit-Nr.: 4007212
Sprache:
Englisch
Umfang: 104 S.
Format in cm: 0.7 x 22 x 15
Einband:
kartoniertes Buch
Erschienen am 12.09.2012
Auflage: 1/2012
- Zusatztext
- 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.
- Autorenportrait
- Juraj Koscak works as Senior Developer commercially.He studied his Master in 2007 and finished Doctorate part-time in 2012from Technical University in Kosice,Department of Cybernetic and Artificial Intelligence, Slovakia.