PAOLA FESTA, University of Napoli FEDERICO II, Italy

GRASP (Greedy Randomized Adaptive Search Procedures) is a multistart metaheuristic for computing good-quality solutions of hard combinatorial optimization problems. Each GRASP iteration is usually made up of a construction phase, where a feasible solution is constructed, and a local search phase which starts at the constructed solution and applies iterative improvement until a locally optimal solution is found. Typically, the construction phase of GRASP is a randomized greedy algorithm, but other types of construction procedures have been also proposed in the literature. Repeated applications of a construction procedure yields diverse starting solutions for the local search and the best overall solution is kept as the output. This talk will give an overview of GRASP enhancements to the basic procedure, including reactive GRASP, intensification strategies, and hybridizations with other metaheuristics. It will be also described a new variant of the GRASP framework that uses a new nonmonotone strategy to explore the neighborhood of the current solution.

Rafael Marti, University of Valencia, Spain

Manuel Laguna, Leeds School of Business of the University of Colorado Boulder. USA

Black box optimizers have a long tradition in the field of operations research. These procedures treat the objective function evaluation as a black box and therefore do not take advantage of its specific structure. Black-box optimization refers to the process in which there is a complete separation between the evaluation of the objective function —and perhaps other functions used to enforce constraints— and the solution procedure. The challenge of optimizing black boxes is to develop methods that can produce outcomes of reasonable quality without taking advantage of problem structure and employing a computational effort that is adequate for the context. Holland’s (1975) genetic algorithms proposal was in fact a black-box optimizer that used an array of bits as the generic representation. The proposed procedure did not include local search and the standard genetic operators (such as single-point crossover) were not linked to the problem context. As GAs became more popular and researchers and practitioner applied them to many hard optimization problems, the context-independent nature of the original proposal began to vanish when improved outcomes were obtained by the addition of problem structure. During the last years we worked on the adaptation of metaheuristic methodologies to deal with black-box combinatorial optimization. We classified the problems according to their solution representation: permutation, binary, and integer to create solvers in each family. In this talk we will review commercial black-box solvers, such as Opttek’s OptQuest, Palisade’s Evolver and Frontline’s Premium Solver, and compare their performance with our metaheuristic implementations based on the Scatter Search methodology on many different optimization problems.

José Andrés Moreno Pérez, Universidad de La Laguna, España

A pesar de su simplicidad, la metaheurística VNS (Búsqueda por Entorno Variable) se ha mostrado muy efectiva en optimización combinatoria y global. Está basada en un cambio sistemático en la estructura de entornos, tanto en la fase de mejora para alcanzar óptimos locales, como en la etapa de agitación para escapar de su atracción. La VNS ha sido aplicada con éxito a una gran cantidad de problemas de optimización con aplicaciones reales, proporcionando soluciones de alta calidad en un tiempo de cómputo razonable. Resulta especialmente indicada para apoyar la toma de decisiones en este ámbito a niveles estratégico, táctico y operativo dado su excelente rendimiento en una gran variedad de problemas de planificación logística. Varios análisis comparativos experimentales han demostrado la superioridad de los algoritmos heurísticos obtenidos utilizando esta joven metaheurística frente a otras alternativas en la solución de problemas estándares de optimización logística. Examinamos los elementos esenciales en la aplicación exitosa de la VNS a los problemas más importantes de optimización logística y se revisan las contribuciones más importantes en esta línea aparecidas en la literatura.

Mauricio Resende, Principal Research Scientist in the transportation area at Amazon, USA.

Nuria Oliver, Data Science Director of Vodafone and the Chief Data Scientist at Data-Pop Alliance.

Ricardo Baeza-Yates, CTO of NTENT, a semantic search technology company based in California, USA.

Henning Wagner, Head of IT Service Management at Rhenus Freight Logistics.

Kristof Roomp, Principal Software Architect working at Microsoft.

Enrique Alfonseca, Staff Research Scientist at Google.