TY - JOUR
T1 - Direct Gravitational Search Algorithm for Global Optimisation Problems
JO - East Asian Journal on Applied Mathematics
VL - 3
SP - 290
EP - 313
PY - 2018
DA - 2018/02
SN - 6
DO - http://doi.org/10.4208/eajam.030915.210416a
UR - https://global-sci.org/intro/article_detail/eajam/10800.html
KW - Gravitational search algorithm, direct search methods, Nelder-Mead method, integer programming problems, minimax problems.
AB - <p style="text-align: justify;">A gravitational search algorithm (GSA) is a meta-heuristic development that
is modelled on the Newtonian law of gravity and mass interaction. Here we propose a
new hybrid algorithm called the Direct Gravitational Search Algorithm (DGSA), which
combines a GSA that can perform a wide exploration and deep exploitation with the
Nelder-Mead method, as a promising direct method capable of an intensification search.
The main drawback of a meta-heuristic algorithm is slow convergence, but in our DGSA
the standard GSA is run for a number of iterations before the best solution obtained is
passed to the Nelder-Mead method to refine it and avoid running iterations that provide
negligible further improvement. We test the DGSA on 7 benchmark integer functions
and 10 benchmark minimax functions to compare the performance against 9 other algorithms, and the numerical results show the optimal or near optimal solution is obtained faster.</p>