Projective dualities for quasiconvex problems (Q2351520)

The paper by Jean-Paul Penot is a valuable contribution to the interfaces between global and nonsmooth optimization, between advances in theory and the foundations of new algorithms. This work is a rigorous study based on which future investigations may be prepared and real-world applications gradually made. Indeed, Jean-Paul Penot studies two dualities that can be applied to quasiconvex problems. These two dualities are conjugacies deduced from polarities; they can be characterized by the polar set of sublevel sets. The author gives some calculus rules for the associated subdifferentials and he relates them with known subdifferentials. He further adapts the general duality schemes in terms of Lagrangians or of perturbations to two particular problems. Firstly, there is a general mathematical programming problem and, secondly, a programming problem with linear constraints are addressed. This excellent article is well-structured, mathematically deep, well demonstrated and exemplified, and written carefully. The five sections of this work are as follows: 1. Introduction, 2. The projective conjugacies, 3. Projective subdifferentials, 4. Perturbations and Lagrangians, and 5. Applications of general mathematical programming problems. These sections comprise the high number of 18 results and definitions - each of them developed, chosen and put with care and consideration, likewise the whole calculus is in this work. Future analytic and algorithmic refinements and extensions, strong mathematical results and optimization methods may be expected in the research community, initiated by this research paper. These may take place in expressions and contexts of Convex, Nonconvex and Generalized Convex Optimization, Semi-Infinite and Infinite Optimization, Optimal Control, Data Mining, Machine Learning, Theory of Inverse Problems, Economics and Game Theory. A further progress like this might serve for achievements in science and technology, industries and economies, earth and environmental sciences, neuroscience and medicine, etc.