An appropriately designed architecture of a neural network is essential to
many realistic pattern-recognition tasks. A choice of just the right number
of neurons, and their interconnections, can cut learning cos ts by orders
of magnitude, and still warrant high classification accuracy. Surprisingly,
textbooks often neglect this issue. A specialist seeking systematic inform
ation will soon realize that relevant material is scattered over diverse so
urces, each with a different perspective, terminology and goals. This brief
survey attempts to rectify the situation by explaining the involved aspect
s, and by describing some of the fundamental techniques.