Search trajectory networks: A tool for analysing and visualising the behaviour of metaheuristics

[EN]A large number of metaheuristics inspired by natural and social phenomena have been proposed inthe last few decades, each trying to be more powerful and innovative than others. However, thereis a lack of accessible tools to analyse, contrast and visualise the behaviour of metaheuristics whensolv...

ver descrição completa

Detalhes bibliográficos
Autores: Ochoa, Gabriela, Malan, Katherine Mary, Blum, Christian
Formato: artículo
Estado:Versión aceptada para publicación
Fecha de publicación:2021
País:España
Recursos:Consejo Superior de Investigaciones Científicas (CSIC)
Repositorio:DIGITAL.CSIC. Repositorio Institucional del CSIC
OAI Identifier:oai:digital.csic.es:10261/240973
Acesso em linha:http://hdl.handle.net/10261/240973
Access Level:acceso abierto
Palavra-chave:Algorithm analysis
Search trajectories
Complex networks
Continuous optimization
Combinatorial optimization
Visualisation
Descrição
Resumo:[EN]A large number of metaheuristics inspired by natural and social phenomena have been proposed inthe last few decades, each trying to be more powerful and innovative than others. However, thereis a lack of accessible tools to analyse, contrast and visualise the behaviour of metaheuristics whensolving optimisation problems. When the metaphors are stripped away, are these algorithms differentin their behaviour? To help to answer this question, we propose a data-driven, graph-based model,search trajectory networks(STNs) in order to analyse, visualise and directly contrast the behaviour ofdifferenttypesofmetaheuristics.Onestrengthofourapproachisthatitdoesnotrequireanyadditionalsampling or algorithmic methods. Instead, the models are constructed from data gathered while themetaheuristics are solving the optimisation problems. We present our methodology, and considerin detail two case studies covering both continuous and combinatorial optimisation. In terms ofmetaheuristics,ourcasestudiescoverthemaincurrentparadigms:evolutionary,swarm,andstochasticlocal search approaches.