Abstract
A wide range of applications use optimization algorithms to find an optimal value, often a minimum one, for a given function. Depending on the application, both the optimization algorithm’s behavior, and its computational time, can prove to be critical issues. In this paper, we present our efficient parallel proposals of the Jaya algorithm, a recent optimization algorithm that enables one to solve constrained and unconstrained optimization problems. We tested parallel Jaya algorithms for shared, distributed, and heterogeneous memory platforms, obtaining good parallel performance while leaving Jaya algorithm behavior unchanged. Parallel performance was analyzed using 30 unconstrained functions reaching a speed-up of up to 57.6 x using 60 processors. For all tested functions, the parallel distributed memory algorithm obtained parallel efficiencies that were nearly ideal, and combining it with the shared memory algorithm allowed us to obtain good parallel performance. The experimental results show a good parallel performance regardless of the nature of the function to be optimized.
Highlights
Optimization algorithms aim at finding an optimal value for a given function within a constrained domain
An analysis of the results revealed that Jaya was superior to, or could compete with, the others when applied to the problems in question
The parallel platform used was composed of 10 HP Proliant SL390 G7 nodes, where each node was equipped with two Intel
Summary
Optimization algorithms aim at finding an optimal value for a given function within a constrained domain. A set of 10 test functions were evaluated when running the algorithm on a single core architecture, and were compared on architectures ranging from 2 to 32 cores They obtain average speed-up values of 4.9x and 6.4x with 16 and 32 processors, respectively. Deep learning is not an optimization algorithm in itself, but the deep network has an objective function, so a heuristic optimization algorithm can be used to tune the network Another important field is data mining that applies to scientific areas [28,29,30]a where Jaya can be applied further; for example, in [31], data optimization techniques and data mining are used together to develop a hybrid optimization algorithm.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have