A new Improved Gravitational Search Algorithm for Function Optimization using a novel 'best-so-far' Update Mechanism

摘要

The focus of this paper is the memory-less Gravitational Search Algorithm (GSA), which is a unique nature inspired algorithms for continuous optimization, based on the laws of gravity and laws of motion. In order to improve the efficiency, reliability and robustness of GSA, an improved GSA is presented in this paper, which incorporates a simple update mechanism of "best-so-far" particle. The performance of Improved GSA and original GSA is well tested on a set of 7 scalable unimodal functions, 6 scalable multi modal functions and 10 non-scalable functions with varying difficulty levels. These 23 problems are the same problems which were presented in the original paper of GSA. Based on the extensive computational analysis it is shown that the improved GSA outperforms original GSA in terms of improved solution quality and faster convergence.