Jian-Wen Peng, Jie Ren, Abhishek Singh, Elisabeth Köbis, A trust-region algorithm with a $T_k$-based nonmonotone strategy for unconstrained multiobjective optimization
Full Text: PDF
DOI: 10.23952/jnva.11.2027.2.08
Volume 11, Issue 2, 1 April 2027, Pages 393-418
Abstract. Trust-region methods provide a reliable framework for solving unconstrained multiobjective optimization problems, but classical monotone acceptance rules may lead to conservative steps and slow progress on challenging problems. This paper presents a trust-region algorithm based on a -type nonmonotone strategy that introduces greater flexibility into the acceptance mechanism while preserving the convergence properties of the method. Under standard assumptions, we prove that every accumulation point generated by the algorithm satisfies the first-order Pareto criticality conditions. The proposed method is tested on a diverse set of benchmark problems and compared with three existing trust-region algorithms. The numerical results demonstrate competitive performance in terms of iterations, objective function evaluations, gradient evaluations, and CPU time, while maintaining high-quality Pareto approximations measured by Hypervolume (HV), Inverted Generational Distance Plus (IGD+), and Purity. Performance profile analysis further confirms the efficiency and robustness of the proposed approach.
How to Cite this Article:
J.W. Peng, J. Ren, A. Singh, E. Köbis, A trust-region algorithm with a -based nonmonotone strategy for unconstrained multiobjective optimization, J. Nonlinear Var. Anal. 11 (2027), 393-418.
