An Inverse Distance Weighting and Radial Basis Function Based Particle Swarm Optimizer for High-dimensional Expensive Optimization

Surrogate-assisted evolutionary algorithms (SAEAs) have emerged as effective approaches for solving highdimensional expensive optimization problems. However, existing surrogate models still face challenges in terms of computational complexity and uncertainty quantification. To address these issues, this paper proposes an inverse distance weighting (IDW) and radial basis function (RBF) based particle swarm optimizer (IR-PSO). The IR-PSO innovatively combines the IDW with RBF models to construct a hybrid IDW-RBF surrogate model. This model leverages the local interpolation properties of IDW and the nonlinear fitting capabilities of RBF to achieve rapid objective function estimation at a low computational cost. Additionally, a simple multi-objective infill criterion (SMIC) is introduced to balance exploration and exploitation. We evaluated IR-PSO on six widely used benchmark functions with dimensions ranging from 50 to 100. The experimental results show that the algorithm has demonstrated competitiveness.