Seed Throwing Optimization: A Probabilistic Technique for Multimodal Function Optimization
Oliver Weede, Alexander Kettler, Heinz Wörn · 2009
A method for optimization of continuous nonlinearfunctions is introduced. Seed Throwing Optimization is aprobabilistic metaheuristic. It has roots in hill climbing and theevolutionary computation like technique harmony search. Therelationship to these algorithms is shown in this paper. Ourmethod is tested in a benchmark and compared to othermetaheuristics. Seed Throwing Optimization is a randomizedgradient ascent with multi initial states and the possibility toexplore only paths which have shown to be good. We alsodeveloped an efficient method for implementing gradientascent without using a gradient.