Surrogate-assisted Evolutionary Algorithm

International Conference on Parallel Problem Solving from Nature (PPSN 2024)

This work proposes a new surrogate-assisted evolutionary algorithm that partially optimizes each objective/constraint, namely surrogate-assisted partial optimization (SAPO). Solutions with better values of objective/constraint are selected from the evaluated solutions as the parent solutions and a focused objective/constraint is independently optimized using surrogate models one by one.

This paper presents a novel surrogate-assisted evolutionary algorithm based on scalarization function approximation, which is designed to strengthen its robustness against this deterioration. The proposed algorithm, called SFA/DE, constructs an approximation model for each scalarization function defined in a decomposition-based framework. Each decomposed problem is then solved using multiple independent models trained for its neighbor problems.

I gave a presentation at JPNSEC Symposium on Evolutionary Computation 2023 (Ohoribata Convention Hall, Odawara, 21-23, Dec.). I was also a student member of organizers for the Open Space Discussion. Thank you to all those who helped make this event possible.

Notice All materials on this page are author’s versions, not necessarily coincide with final published versions.

Notice All materials on this page are author’s versions, not necessarily coincide with final published versions.

JPNSEC Symposium on Evolutionary Computation 2023

The Genetic and Evolutionary Computation Conference (GECCO 2023)

I gave a poster presentation at The Genetic and Evolutionary Computation Conference (GECCO 2023, Altis Grand Hotel, Lisbon, Portugal, 15-19, Jul). Thank you to all those who helped make this event possible.

This paper utilizes the expected gradient of the Gaussian Process (GP) in a surrogate-assisted evolutionary algorithm. Specifically, our proposal iteratively runs a quasi-Newton method (L-BFGS-B) changing initial points on multiple GPs constructed to approximate the promising region of the objective function.