@article {pub4405pub4705,
	title = {Transfer Learning Based Surrogate Assisted
Evolutionary Bi-objective Optimization for Objectives with Non-uniform Evaluation Times},
	author = {Xilu Wang AND Yaochu Jin AND Sebastian Schmitt AND Markus Olhofer AND Richard Allmendinger},
	year = {2021},
	abstract = {Various multiobjective optimization algorithms have
been proposed with a common assumption that the evaluation of
each objective function takes the same period of time. Little
attention is paid to more general and realistic optimization
scenarios where different objectives are evaluated by different
computer simulations or physical experiments with different
time complexity. To address this issue, we investigate benchmark
scenarios with two objectives where one objective evaluation
is assumed to be much slower than the other. We propose a
transfer learning scheme within a surrogate-assisted evolutionary
algorithm framework (Tr-SAEA) to augment data for the slow
objective by transferring knowledge between objectives. To this
end, a hybrid domain adaptation method aligning the second-
order statistics and marginal distributions across domains is
introduced to generate promising samples for the slow objective
according to the search experience of the fast one. To make
use of these unlabeled samples, a Gaussian process (GP) based
co-training method is proposed to select newly labeled data with
regard to its confidence level and further boost each GP regressor
accuracy and enhance the convergence. Our experimental results
on three test suites demonstrate that the proposed algorithm is
competitive for solving bi-objective optimization problems where
objectives have non-uniform evaluation times, compared with the
state-of-the-art delay-handling methods.},
	publisher = {Elsevier},
	journal = {Knowledge-Based Systems},
	volume = {227},
	number = {5},
	pages = {107190}
}