A proximal-point outer approximation algorithm
| Autor: | Goele Pipeleers, Jan Swevers, Joris Gillis, Massimo De Mauri |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Mathematical optimization
021103 operations research Control and Optimization Applied Mathematics 0211 other engineering and technologies Regular polygon Approximation algorithm 010103 numerical & computational mathematics 02 engineering and technology 01 natural sciences Execution time Scheduling (computing) Proximal point Computational Mathematics Hybrid system 0101 mathematics Heuristics Implementation Mathematics |
| Zdroj: | Computational Optimization and Applications. 77:755-777 |
| ISSN: | 1573-2894 0926-6003 |
| DOI: | 10.1007/s10589-020-00216-9 |
| Popis: | Many engineering and scientific applications, e.g. resource allocation, control of hybrid systems, scheduling, etc., require the solution of mixed-integer non-linear problems (MINLPs). Problems of such class combine the high computational burden arising from considering discrete variables with the complexity of non-linear functions. As a consequence, the development of algorithms able to efficiently solve medium-large MINLPs is still an interesting field of research. In the last decades, several approaches to tackle MINLPs have been developed. Some of such approaches, usually defined as exact methods, aim at finding a globally optimal solution for a given MINLP at expense of a long execution time, while others, generally defined as heuristics, aim at discovering suboptimal feasible solutions in the shortest time possible. Among the various proposed paradigms, outer approximation (OA) and feasibility pump (FP), respectively as exact method and as heuristic, deserve a special mention for the number of relevant publications and successful implementations related to them. In this paper we present a new exact method for convex mixed-integer non-linear programming called proximal outer approximation (POA). POA blends the fundamental ideas behind FP into the general OA scheme that attepts to yield faster and more robust convergence with respect to OA while retaining the good performances in terms of fast generation of feasible solutions of FP. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|