Abstract:
Based on observations from nature and promising experimental data, this work opens a new direction in the field of evolutionary computation: the usage of disagreements to derive superior algorithms with enhanced exploitation and exploration capabilities for global optimization problems. Initially, good results are obtained with the new 6σ–PSOD operator which imitates the natural distribution of the disagreements in nature with a low computational cost. The disagreements are applied only to the social component from PSO. The partial disagreements increase the exploitation, while the extreme disagreements enhance the exploration. Disagreements can also be used to mitigate stagnation in swarm with the new RS–PSOD operator. Further good results with the concept of disagreements on real-valued genetic algorithms open the door for experimenting with this new idea in the larger field of the evolutionary computation. This thesis is a proof of concept for disagreements in swarm intelligence, through PSO, and in the larger category of the evolutionary computation, through real-valued genetic algorithms. Because its contributions are at the core of the algorithms in study, the potential impact can positively affect all areas in which evolutionary computation is used.
