Random approximations in multiobjective optimization (Q2364484)

The article is a valuable contribution to optimization theory right at the interface between multi-criteria decision making and stochastic programing. This mathematical work is rigorous and well-organized. Based on it, future research can be designed and established, and real-life applications made, e.g., portfolio optimization from finance and actuarial sciences, in time and under regime switching and paradigm shifting, quality control in production and service sectors, decision making in social sciences, environmental and developmental sciences, etc. In fact, decision makers often cope with programming problems containing unknown quantities. They estimate these quantities, and solve the program right as it appears then -- the approximate problem. Hence, there is given a necessity to introduce conditions which will guaranteed that the solutions to the approximate problem come close enough to the true problem's solutions in a meaningful manner. Confidence sets, i.e., regions which cover the true sets under a certain probability, give useful quantitative information. The author studies multi-objective problems, and for the sets of efficient points and of weakly efficient points, as well as for the corresponding solution sets, she derives confidence sets. In addition to the important convergence conditions for objective or constraint functions, one approach for obtaining confidence sets needs some a priori knowledge about the true problem which, however, may not be given. For this reason, another so-called relaxation method is proposed. In fact, this one works without any a priori information regarding the true problem. The author applies the results on the famous mean-variance approach of portfolio optimization, called as the Markowitz model. This excellent paper is well-embedded into the scientific landscape, including future potentials, well exemplified and illustrated, and well written. The six sections of this deep and valuable work are as follows: 1. Introduction, 2. Mathematical model, consisting of: 2.1 General model, and 2.2 The Markovitz model, 3. Approximation of the image set, 4. Approximation of the sets of efficient points and the sets of weakly efficient points, 5. Approximation of the solution sets, 6. Relaxation, consisting of: 6.1 Relaxation with partial knowledge, 6.2 Relaxation without knowledge about the true model, and 6.3 Application to the Markowitz model. Further refinements and improvements in approaches, methods and implementations, results and treatments can be expected in the scientific community, inspired and prepared by this research paper. Those further contributions could be made by addressing (i) various forms of uncertainty and its management (including robust optimization, robust portfolio optimization, e.g., under parallelepiped uncertainty, stochastic calculus, risk management, optimal control of expected utility, of processes with jumps and regime switching, etc.), of the environment, and of the ``human factor'', (iii) various forms of involving the time (discrete and continuous time, change-of-time method, etc.), and various other forms of multi-objective optimization (e.g., optimization and robust optimization of so-called desirability functions, Tikhonov regularization and ridge regression, MARS, CMARS, RMARS, RCMARS, etc.). Such future advances can stimulate progress in natural sciences, in engineering and medicine, in bio-, earth- and environmental sciences, in neuroscience, healthcare and social sciences, in economics and game theory, in computer engineering, in image, speech and video processing.