Knowledge Transfer in Data-Driven Evolutionary Optimization

Abstract

Lack of training data is one major challenge in data-driven optimization, since data collection is either computationally expensive or costly in many data-driven optimization problems. To address this issue, this chapter presents three classes of knowledge transfer approaches in data-driven evolutionary optimization. The first approach is based on semi-supervised learning, transferring knowledge from unlabeled data to labeled data. The second approach makes use of transfer learning with the help of parameter sharing and domain adaptation, to transfer knowledge between objectives or problems. Finally, transfer optimization, a variant of multi-tasking optimization, is employed to transfer knowledge between multi-fidelity formulation or multi-scenarios of the same optimization problem.