Robust Parallel Genetic Algorithms with Re-initialisation

The influence of different parallel genetic algorithm (PGA) architectures on the GA convergence properties is analysed. Next, two proposed versions of these PGA architectures are compared – homogenous and heterogeneous. Finally the effect of re-initialisation in some partial populations on the PGA convergence has been analysed. The proposed PGA modifications are useful mainly in case of non-smooth cost function optimisation.

• Type: Conference
• Year: 2004
• Where: PPSN
• Authors: Ivan Sekaj
• Keywords: Distributed And Parallel Computing, GA Convergence Properties, Parallel Genetic Algorithm, PGA Architectures, PPSN 2004