The imbalances and conditioning of the objective functions influence the performance of first-order methods for multiobjective optimization problems (MOPs).
To balance per-iteration cost and better curvature exploration, a Barzilai-Borwein descent method with variable metrics (BBDMO_VM) is proposed.
BBDMO_VM employs a variable metric in the direction-finding subproblems to explore the curvature of all objectives and mitigate the effect of imbalances.
Comparative numerical results confirm the efficiency of the proposed BBDMO_VM method for large-scale and ill-conditioned multiobjective optimization problems.
A nonmonotone proximal quasi-Newton method for unconstrained convex multiobjective composite optimization problems is proposed.
The method employs a nonmonotone line search and converges to a Pareto optimal solution.
Numerical experiments show the effectiveness of the proposed algorithm on a set of test problems.