Title :
Co-operating Populations with Different Evolution Behaviours
Author :
Adamidis, P. ; Petridis, V.
Author_Institution :
Dept. of Electr. & Comput. Eng., Aristotelian Univ. of Thessaloniki, Greece
Abstract :
Parallel genetic algorithms (PGA) offer a natural and productive way to solve a problem better than a single population. Up to now the different PGA paradigms use the same evolution behaviour on each population. This paper proposes a method, called Co-operating Populations with Different Evolution Behaviours (CoPDEB) where the populations are allowed to exhibit different evolution behaviours. This is achieved by using a variety of selection mechanisms, operators, communication methods and parameters as it is explained in the sequel. This method has been tested on the problem of training a recurrent artificial neural network (RANN)
Keywords :
biocybernetics; cooperative systems; evolution (biological); genetic algorithms; learning (artificial intelligence); parallel algorithms; problem solving; recurrent neural nets; CoPDEB; Cooperating Populations with Different Evolution Behaviours; communication methods; evolution behaviour; operators; parallel genetic algorithms; problem solving; recurrent artificial neural network training; selection mechanisms; single population; Convergence; Electronics packaging; Frequency; Genetic mutations; Network topology; Neural networks; Steady-state; Wheels;
Conference_Titel :
Evolutionary Computation, 1996., Proceedings of IEEE International Conference on
Conference_Location :
Nagoya
Print_ISBN :
0-7803-2902-3
DOI :
10.1109/ICEC.1996.542358
