Hill-climbing, simulated annealing and genetic algorithms: a comparative study and application to the mapping problem

Abstract

INTRODUCTIONHill-climbing, simulated annealing and genetic algorithms are search techniques that can be applied to most combinatorial optimization problems. In this paper, the three algorithms are used to solve the mapping problem: optimal static allocation of Communicating processes (tasks, objects, agents) on distributed memory architectures.Each algorithm is independently evaluated and optimized according to its parameters. The parallelization of the algorithms is also considered. As an example, a massively parallel genetic algorithm is proposed for the problem, and results of its implementation on a 128-processor Supemode (reconfigurable network of transputers) are given.A comparative study of the algorithms is then carried out. The criteria of performances considered are the quality of the solutions obtained and the amount of search timeIn this paper, we are interested to the mapping problem: optimal static placement of communicating processes on the processors of a distributed memory parallel machine.The problem is known to be NP-complete [Garey79].Consequently, heuristic methods shall be used. They may find only approximations of the optimum, but they will do it in a "reasonable" amount of time.Heuristic algorithms may be divided in two main classes. First, the general purpose optimization algorithms independent of the given optimization problem and, on the other hand, the heuristic approaches especially designed for the mapping problem. As we want to avoid the intrinsic disadvantd,: of the algorithms of this second class (their limited applicability), this paper is only concerned with the first class of algorithms. used for several benchmarks. A hybrid approach consisting in a combination of genetic algorithms and hill-climbing is also proposed and evaluated.

Arrangement of multi-dimensional scalable video data for heterogeneous clients

“…There have been several meta-heuristic algorithms that attempt to quickly find near-optimal solutions for combinatorial optimization problems such as our formulated problem. Since it is not guaranteed that the search space of the proposed formulated problem is convex space, we employ the simulated annealing (SA) algorithm which can avoid the problem of getting struck on a local minimum [30,33,34].…”

A Two-Phase Mapping Method for Parallel Applications Simulation

“…Obtaining the optimal partition of graph G is known to be NP-complete. Researchers have proposed many heuristic algorithms to get approximate solutions [11], which is still time-consuming. Mapping methods based on array assignment, also known as block-cyclic mapping methods perform well in balancing load and minimizing communication for regular parallel programs [12].…”

Job Scheduling Algorithms on Grid Computing: State-of- the Art

“…HC is a local search mechanism which is suitable for finding local optimal solutions and so it is not appropriate in searching for global optimization. Besides, HC optimization mechanism suffers from the plateau problem when the solution search space is flat; in that situation, HC is not capable of finding out which way it should go, or it may choose directions that never lead to the optimal solution [42].…”