Withering process for tree-seed algorithm

The tree-seed algorithm (TSA) is first proposed to solve continuous optimization problems in 2015. The possible solutions of the optimization problems are represented as trees and seeds. Being used the relation between trees and seeds, the candidate solutions are obtained during the iterations. While a seed which is a possible solution is created, either best solution or current tree is used in updating strategy of TSA and the usage of best or current tree is controlled by a control parameter named as search tendency. For especially multimodal numeric benchmark functions, the population goes into stagnation behavior after a while because the stand swarms to a place on the solution space. Therefore, the new candidate solutions are very similar to previous solutions. To overcome this issue of TSA, a new control parameter named as withering process (WP) is added to basic version of TSA. The novel proposed version of TSA is called as TSAWP and applied to solve 13 multimodal benchmark functions. Experimental results show that TSAWP is better than the basic version of TSA in terms of solution quality and robustness based on standard deviations in most cases.