An application oriented comparison of optimization and neural network based design techniques

Abstract

This paper investigates the use of Artificial Neural Networks compared to conventional optimization techniques in the area of VLSI placement. Two different placement methods are considered, and, although initially it seems that the conventional methods described are faster, when more complex layouts with no fixed points are encountered, the self-organisation network produces better results.

Fast index assignment algorithm for vector quantisationover noisy transmission channels

“…The index assignment process, where a code book is re-ordered such that similar reference vectors are represented by indices with similar binary patterns, is analogous to the problem of placement in VLSI circuits, where modules are placed on a two-dimensional surface so as to minimise the total wire length [3]. The aim of Index Assignment can be viewed as placing the n = 2 r reference vectors in the r-dimensional binary space described by the codewords, such that similar reference vectors are located in similar locations.…”