An EA framework for uncertain optimization problem
Many engineering design projects involve solving complex optimization problems. Evolutionary algorithms (EA) have been widely accepted as efficient optimizers for complex real life problems. However, many real life optimization problems involve time-variant noisy environment, which pose major challenges to EA-based optimization. Presence of noise interferes with the evaluation and the selection process of EA and adversely affects the performance of the algorithm [7]. Also presence of noise means fitness function can not be evaluated and it has to be estimated instead. Several approaches have been tried to overcome this problem, such as introduction of diversity (hyper mutation, random immigrants, special operators) or incorporation of memory of the past (diploidy, case based memory) [8]. In this paper we propose a method, DPGA (distributed population genetic algorithm) that uses a distributed population based architecture to simulate a distributed, self-adaptive memory of the solution space. Local regression is used in each sub-population to estimate the fitness. Specific problem category considered is that of optimization of functions with time variant noisy fitness. Successful applications to benchmark test problems ascertain the proposed method’s superior performance in terms of both adaptability and accuracy.
Dear author,
Can you please provide more information on how candidates for pseudo-populations are selected?
Thank you.
D Pham.
"Next the regenerated main population dissolves into
pseudo-populations by self-organization. This is
essentially distribution of the candidate solutions
into pseudo-populations based on specific criteria.
A factor of fitness and population size decides the
eligibility of a pseudo-population to obtain
evolution right. An eligible pseudo-population then
evolves by the canonical GA mechanism. Local
regression is used to estimate the fitness values.
Mutation rate depends on a factor of fitness and
population size as well."