Grey Wolf Optimizer: Population Search, Tuning, and Benchmark Limits

Summary

The Grey Wolf Optimizer is presented as a population-based stochastic method for optimizing a fitness function. Candidate solutions are ranked as alpha, beta, and delta leaders, while the remaining candidates update their positions using those leaders. A parameter that decreases over iterations shifts the search from broad exploration toward local convergence; random coefficients affect the updates. The article also describes an implementation that allows the number of leaders to vary, sorts candidates by fitness, and supports bounded parameter ranges and steps.

The article compares the optimizer with other population methods and random search on smooth and discrete test problems, reporting results across problems of different dimensionality. It characterizes GWO as fast and effective on smooth functions with many variables, while cautioning that it can settle in local optima and scales poorly on discrete or nondifferentiable functions. These benchmark outcomes concern the tested functions and configurations; they do not establish trading performance or guarantee that the method will work well on a particular strategy optimization problem.

Key ideas

• The algorithm ranks candidate solutions and uses the three strongest as guides for the rest of the population.
• A shrinking control parameter moves the search from exploration toward convergence.
• The implementation permits a configurable number of leaders and bounded parameter values.
• Reported comparisons cover test functions, including smooth and discrete cases.
• The article warns of local optima and weak scaling on discrete or nondifferentiable problems.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.