» Investigación » SISIUS - Investigación en la USE

Responsable: Emilio Carrizosa Priego
Tipo de Proyecto/Ayuda: Plan Nacional del 2009
Referencia: MTM2009-14039-C06-06
Fecha de Inicio: 01-01-2010
Fecha de Finalización: 31-12-2013

Empresa/Organismo financiador/es:

• Ministerio de Ciencia e Innovación

Equipo:

• Investigadores:
  • Silvia Bermúdez Parrado
  • Rafael Blanquero Bravo
  • Belén Martín Barragán
  • Alba Victoria Olivares Nadal (alta: 01/06/2012)
  • Frank Plastria
  • María Dolores Romero Morales
• Personal Investigador en Formación:
  • Amaya Nogales Gómez (alta: 01/10/2010)

Contratados:

• Técnicos/Personal Administrativo:
  • Vanesa Guerrero Lozano
  • Josefa Ramírez Cobo

Resumen del proyecto:

This project addresses several issues related with Mathematical Optimization and its applications in Operations Research and Statistics. 

Global-Optimization methods are being designed for DC functions (functions which can be expressed as the difference of two convex functions). How to obtain sharp bounds, and how to successfully use such bounds to speed convergence of branch-and-bound algorithms is under investigation. Applications to several problems in Continuous Location and neighboring fields are being explored. 

For problems of larger dimensions, branch and bound methods are too demanding in terms of running times. Instead, heuristic methods must be used. Global-search strategies, such as Variable Neighborhood search, combined with structural properties of the functions and problems, are being analyzed.

Data Mining, and more precisely, Classification, is another source of challenging optimization problems. The research team is addressing Support-Vector-Machines-based strategies for Supervised Classification. Modifying the objective function or asking also for feature selection transforms the standard SVM into a challenging global optimization problem, for which exact and heuristic approaches are being designed and numerically tested. Unsupervised classification problems are also investigated. In particular, heuristics for clustering problems with alternative objective functions are being derived.