Probabilistic Selection Approaches in Decomposition-Based Evolutionary Algorithms for Offline Data-Driven Multiobjective Optimization

In offline data-driven multiobjective optimization, no new data are available during the optimization process. Approximation models, also known as surrogates, built using provided data. A evolutionary algorithm can be utilized to find solutions by these surrogates. The accuracy of approximated depends on surrogates and approximations typically involve uncertainties. this article, we propose probabilistic selection approaches that utilize uncertainty information Kriging models (as surrogates) improve solution process in optimization. These designed for decomposition-based algorithms can, thus, handle a large number objectives. proposed were tested distance-based visualizable test problems DTLZ suite. produced with greater hypervolume, lower root mean squared error compared generic transfer learning approach do not use information.