Can I read Training Nets of Stochastic Units Using System Identification on EtoBox?

Training Nets of Stochastic Units Using System Identification by Kevin Gurney is a Computer Science article available to read on EtoBox.

What is Training Nets of Stochastic Units Using System Identification about?

Some of the techniques of System Ident~/ication ( SID ) in control t/woO' have been used to train.l~'ed/brward nets of units which use pulse fi'equency coding to communicate their output. The algorithm makes use o./a vector of signals at each training step but. unlike back error propagation, these are broa&'ast to the net as a whole. This paper investigates the nature of training in two types of net. One uses a stochastic variant of the semilinear Threshold Logic Unit (TLU), while the other uses digital units &:l~ned by populations of vahws at the vertices of the n-cube. An investigation is made into the effect of training parameters on learning in both cases, leading to a comparison of the two net types on a single problem (the 838 encoder). The greater fimctionalio, of the cubic nodes leads to reduced training times and a d~ff'erence in behavior under change of training parameters.

Who reads Training Nets of Stochastic Units Using System Identification?

Author

Kevin Gurney

Publisher

Elsevier Science; Elsevier ; Elsevier Ltd.; Elsevier BV (ISSN 0893-6080)

Published

1993

Language

EN

Field

Computer Science (Physical Sciences)