On optimal static and dynamic parameter choices for fixed-target optimization

Dmitry Vinokurov, Maxim Buzdalov

Allbwn ymchwil: Pennod mewn Llyfr/Adroddiad/Trafodion CynhadleddTrafodion Cynhadledd (Nid-Cyfnodolyn fathau)

Crynodeb

Various flavours of parameter setting, such as (static) parameter tuning and (dynamic) parameter control, receive a lot of attention both in applications and in theoretical investigations. It is widely acknowledged that the choice of parameter values influences the efficiency of evolutionary algorithms (and other search heuristics) a lot, and a considerable amount of work has been dedicated to finding (near-)optimal choices of parameters, or of parameter control strategies. It is perhaps surprising that all the recent theoretic attempts aim at making smaller the time needed to reach the optimum, whereas in most practical settings we may only hope to reach certain realistic fitness targets. One of the ways to close this gap is to study static and dynamic parameter choices in fixed-target settings, and to understand how these choices are different from those tuned towards reaching the optimum. In this paper we investigate some of these settings, using a mixture of exact theory-driven computations and experimental evaluation, and find few remarkably generic trends, some of which may explain a number of misconceptions found in evolutionary computation.

Iaith wreiddiolSaesneg
TeitlProceedings of Genetic and Evolutionary Computation Conference
CyhoeddwrAssociation for Computing Machinery, Inc
Tudalennau876-883
Nifer y tudalennau8
ISBN (Electronig)9781450392372
Dynodwyr Gwrthrych Digidol (DOIs)
StatwsCyhoeddwyd - 08 Gorff 2022
Digwyddiad2022 Genetic and Evolutionary Computation Conference, GECCO 2022 - Virtual, Online, Unol Daleithiau America
Hyd: 09 Gorff 202213 Gorff 2022

Cyfres gyhoeddiadau

EnwGECCO 2022 - Proceedings of the 2022 Genetic and Evolutionary Computation Conference

Cynhadledd

Cynhadledd2022 Genetic and Evolutionary Computation Conference, GECCO 2022
Gwlad/TiriogaethUnol Daleithiau America
DinasVirtual, Online
Cyfnod09 Gorff 202213 Gorff 2022

Ôl bys

Gweld gwybodaeth am bynciau ymchwil 'On optimal static and dynamic parameter choices for fixed-target optimization'. Gyda’i gilydd, maen nhw’n ffurfio ôl bys unigryw.

Dyfynnu hyn