abstract = "Many-objective optimization problems (MaOPs) present
unique challenges, particularly in achieving both
convergence and diversity across high-dimensional
Pareto fronts. These challenges become even more
pronounced when the Pareto fronts are irregular or
disconnected. Traditional methods that rely on
uniformly distributed reference points often struggle
to perform well under such conditions. This paper
introduces a new algorithm called Model Adaptive
Reference Points NSGA-III (MARP-NSGA-III) to address
these issues. The algorithm uses Gaussian Processes
(GPs) to generate reference points based on the dynamic
distribution of solutions. By incorporating NSGA-III,
MARP-NSGA-III models the evolving Pareto front and
adaptively allocates reference points, ensuring they
align more closely with the front's structure. This
study provides a detailed framework explaining the core
components of the method, including how the GP model is
constructed, how areas of interest are estimated, and
how reference points are generated. MARP-NSGA-III
represents a significant step forward in solving 11
MaOPs, offering a flexible and adaptive approach. Our
experiments show that, for benchmark problems, our
proposed algorithm (MARP-NSGAIII) performs better than
other many-objective competing algorithms. Gaussian
Process-Driven Adaptive Reference Point Generation for
Many-Objective Optimization.",
notes = "GECCO-2025 EMO A Recombination of the 34th
International Conference on Genetic Algorithms (ICGA)
and the 30th Annual Genetic Programming Conference
(GP)