GPU implementation of TLBO algorithm to test constrained and unconstrained benchmark functions
Shah Sujeeta Ramanlal, Shinde Santaji Krishna · 2016
The objective function may have many local optima whereas the designer is interested in the global optimum. The classical methods (e.g. Gradient methods) were cannot be handled by such problems. They only compute the local optima. So there remains a need for efficient and effective optimization methods for mechanical design problems. The Continuous research is being conducted in this field by different optimization techniques. The nature-inspired heuristic optimization methods are proving to be better than deterministic methods. In this paper, the Performance of algorithm on unconstrained and constrained benchmark functions is tested by running these algorithms sequentially. The main motivation of using parallel computing is to improve the performance of these algorithms parallely. GPGPU is applicable where data parallelism and independency is possible, a good implementation on a GPGPU can achieve more than 100 times better speedup over sequential execution. CUDA implementation proves better in the evolutionary algorithm not only regarding of speed up but also in convergence time.