A recursive random search algorithm for network parameter optimization

摘要: This paper proposes a new heuristic search algorithm, Recursive Random Search(RRS), for black-box optimization problems. Specifically, this algorithm is designed the dynamical parameter of network protocols which emphasizes on obtaining good solutions within limited time frame rather than full optimization. The RRS based initial high-efficiency property random sampling and attempts to maintain by constantly "restarting" with adjusted sample spaces. Due its basis sampling, robust effect noises in objective function it performs especially efficiently when handling functions negligible parameters. These properties have been demonstrated tests suite benchmark functions. has successfully applied optimal configuration several protocols. One application routing presented.