Abstract—This paper discusses novel dedicated hardware architecture for hybrid optimization based on Genetic algorithm (GA) and Simulated Annealing (SA). The proposed architecture achieves high speed processing. Moreover, it achieves the searching not only globally, but also locally. To keep general purpose, self-control processing by a handshake system is introduced. By adopting the handshake system, the proposed architecture can be applied to various combinatorial
optimization problems by only changing an encoder, a decoder, and an evaluation circuit. Furthermore, the proposed architecture realizes flexibility for many genetic operations on
GA. In order to evaluate the proposed architecture, we conduct two kinds of experiments. One is an experiment which applies the proposed architecture to TSP, and the other is an experiment which applies it to VLSI floorplanning. These experiment results prove that the proposed architecture achieves high speed processing, while keeping the quality of the solutions.
Index Terms—Genetic algorithm, Simulated annealing,
Dedicated hardware, General-purpose properties
I. INTRODUCTION
Genetic Algorithm (GA)[1] was proposed by Holland as
an algorithm for probabilistic search, learning, and
optimization, and is based in part on the mechanism of
biological evolution and Darwin’s theory of evolution. This
algorithm is a powerful search tool, particularly when applied
for combinatorial optimization problems[2]-[7]. However,
the implementation of an efficient GA often faces two major
problems, on one side, the premature convergence to local
optima and on the other the requirements for the GA search
of long times in order to reach an optimal or a good
suboptimal solution.
In order to prevent the premature convergence, the
coupling of GA and one point search algorithm (local search
algorithm), such as Simulated Annealing (SA)[8]-[11], to
form hybrid GA can be advantageous. SA repeatedly
generates succeeding solutions using the local search
procedure. Some of them are accepted and some will be
rejected, according to a predefined acceptance rule. The
acceptance rule is motivated by an analogy with annealing
processes in metallurgy as shown in Fig.1 (a).
On the other hand, GA repeatedly propagates generation
đang được dịch, vui lòng đợi..
