Crosshair Optimizer

A novel local search based optimization algorithm named Crosshair Optimizer (CHO) is introduced in this paper. In most algorithms, much of the computational resources are used to explore around a solution space, but in reality, much of the time there are only parts of an optimal solution that can be improved. Using this information, if an algorithm explores the space around an optimal solution using a random number of its dimensions to dictate their placements, rather than all of them, the solution space is explored in a less random fashion. CHO employs a rapid neighbourhood generation on each iteration and selects a sub-sequences of best performing solutions. These selected solutions then randomly generate neighboring solutions only in certain search space dimensions. Exploration is done by randomly generating solutions in only two dimensional axis to the neighbourhood cluster. This speeds up the search process with fine grain sampling, and is quickly able to migrate the search space to another location without using drift migration. Experimentation was conducted on the standard unimodal and multimodal problems, with CHO performing extremely well against standard evolutionary algorithms.