A heuristic to determine the initial gravitational constant of the GSA

Published: 12 November 2018| Version 2 | DOI: 10.17632/nphcty6yp8.2
Contributor:
Alfredo Barbosa

Description

These files are the results of the experiments cited in the paper A heuristic to determine the initial gravitational constant of the GSA, which presentes the gravitational search algorithm with normalized gravitational constant (GSA-NGC). Thirteen benchmark functions were used in these experiments, seven unimodal functions (1-7) and six multimodal functions (8-13): 1. Sphere 2. Schwefel 2.22 3. Schwefel 1.2 4. Schwefel 2.21 5. Rosenbrock 6. Step 7. Quartic 8. Schwefel 2.26 9. Rastrigin 10. Ackley 11. Griewank 12. Penalized 1 13. Penalized 2 Each file is the result of 30 runs of a gravitational search algorithm on a benchmark function. Five series of experiments were conducted: * Experiments varying the initial gravitational constant of GSA (SGSA G0) in the values 1, 10, 100, 1000 and 10000. * Experiments varying the constant beta of the GSA-NGC (GSA-NGC Beta) in the values 0.125, 0.25, 0.5, 1, 2, 4 and 8. * Experiments comparing SGSA and GSA-NGC in a relatively small search space (Search Space Small), each dimension of which is 100 times smaller than the original; * Experiments comparing SGSA and GSA-NGC in a relatively big search space (Search Space Big), each dimension of which is 100 times bigger than the original; and * Experiments comparing SGSA and GSA-NGC in a irregular search space (Search Space Irregular), each dimension of which is a power of ten times the original. The parameters used in these experiments were: number of iterations T=1000, number of particles P=50 (for the first four series) and P=31 (for the last series), number of dimensions D=30 (for the first four series) and D=11 (for the last series), with alpha=20 and using the elitist strategy. These results are also shown in the pdfs. These experiments are further described in the paper. All results can be loaded in Octave using the load function.

Files

Institutions

Universidade Federal de Itajuba

Categories

Artificial Intelligence, Global Optimization

Licence