Participatory search learning algorithms and applications
Yi Ling Liu · 2016
Search is one of the most useful procedures employed in numerous situations such as optimization, machine learning, information processing and retrieval.This work introduces participatory search, a population-based search algorithm based on the participatory learning paradigm.Participatory search is an algorithm in which search progresses forming pools of compatible individuals, keeping the one that is the most compatible with the current best individual in the current population, and introducing random individuals in each algorithm step.Recombination is a convex combination modulated by the compatibility between individuals while mutation is an instance of differential variation modulated by compatibility between selected and recombined individuals.The nature of the recombination and mutation operators are studied.Convergence analysis of the algorithm is pursued within the framework of random search theory.The participatory search algorithm with arithmetical-like recombination is evaluated using ten benchmark real-valued optimization problems, and its performance is compared against population-based optimization algorithms representative of the current state of the art in the area.The participatory search algorithm arithmetical recombination is also evaluated using a suite of twenty eight benchmark functions of the evolutionary, real-valued optimization competition of the IEEE CEC 2013 (IEEE Congress on Evolutionary Computation) to compare its performance against the competition winners.Computational results suggest that the participatory search algorithm is as good as the winners.An application concerning development of fuzzy rule-based models from actual data is given.The performance of the models produced by participatory search algorithms are compared with a state of the art genetic fuzzy system approach.Experimental results suggest that the participatory search algorithm with arithmetical-like recombination performs best.