A federated data-driven evolutionary algorithm for expensive multi-/many-objective optimization

Abstract Data-driven optimization has found many successful applications in the real world and received increased attention field of evolutionary optimization. Most existing algorithms assume that data used for are always available on a central server construction surrogates. This assumption, however, may fail to hold when must be collected distributed way subject privacy restrictions. paper aims propose federated data-driven multi-/many-objective algorithm. To this end, we leverage learning surrogate so multiple clients collaboratively train radial-basis-function-network as global surrogate. Then new acquisition function is proposed approximate objective values using estimate uncertainty level approximated based local models. The performance algorithm verified series benchmark problems by comparing it with two state-of-the-art surrogate-assisted multi-objective algorithms.