文档介绍：WORK
3 ARCHITECTURES
Hooman Yousefizadeh and Ali Zilouchian
INTRODUCTION
Interest in the study of works has grown remarkably in the last two
decades. This is due to the conceptual viewpoint regarding the human brain as a
model of a putation device, a very different one from a traditional
puter. works monly classified by work
topology, node characteristics, learning, or training algorithms. On the other
hand, the potential benefits of works extend beyond the high
computation rates provided by massive parallelism of works. They
typically provide a greater degree of robustness or fault tolerance than Von
Neumann puters. Additionally, adaptation and continuous
learning are ponents of NN. These properties are very beneficial
in areas where the training data sets are limited or the processes are highly
nonlinear. Furthermore, designing artificial works to solve problems
and studying real works (Chapter 4) may also change the way we
think about the problems and may lead us to new insights and algorithm
improvements.
The main goal of this chapter is to provide the readers with the conceptual
overviews of several work architectures. The chapter will not delve
too deeply into the theoretical considerations of any work, but will
concentrate on the mechanism of their operation. Examples are provided for
work to clarify the described algorithms and demonstrate the reliability
of work. In the following four chapters various applications pertaining to
works will be discussed.
This chapter anized as follows. In section , various classifications of
works according to their operations and/or structures are presented.
Feedforward and works are discussed. Furthermore, two different
methods of training, namely supervised and unsupervised learning, are
described. Section is devoted to error back propagation (BP) algorithm.
Various properties of work are also discussed in this section. Radial
basis