2014-05-23: Is CPLEX like a box of chocolates?
É com muito prazer que vos convidamos a participarem no curso
"Is CPLEX like a box of chocolates?"
que será leccionado por Ricardo M. Lima que é actualmente Marie Curie Fellow no Laboratório Nacional de Energia e Geologia.
O curso decorrerá no próximo dia 4 de Junho (nova data), na Faculdade de Ciências e Tecnologia (UNL) terá a seguinte programação:
14:00 – 14:30
14:30 – 16:30
Curso – parte I
16:30 – 17:00
17:00 – 18:00
Curso – parte II
Sendo uma organização conjunta entre o Centro de Matemática e Aplicações e a APDIO, o custo de participação é de 7€ para sócios da APDIO e 10€ para não sócios.
O pagamento deverá ser efectuado por transferência bancária para a conta indicada na ficha de inscrição. Envie, por favor, o comprovativo da transferência, onde deverá contar o nome e número de sócio da APDIO, para o mesmo endereço electrónico (email@example.com) afim de poder ser emitido o correspondente recibo.
This short course addresses some of the recent features and capabilities of the optimization software package CPLEX, which have been contributing for the solution of increasing size problems. In the last years CPLEX has incorporated several developments at the level of solving linear programming (LP) problems, pre-processing, cutting planes, heuristics, and auxiliary tools. The presentation starts with an overview of the CPLEX history, and the type of the problems it can solve. The main focus is the solution of mixed integer linear programming (MILP) problems, and the discussion of both established and new methods used in the branch and cut (B&C) algorithm to solve MILP problems. Examples of new methods are heuristics such as Local Branching (LB), Relaxation Induced Neighborhood Search (RINS) and evolutionary algorithms for polishing MILP solutions. These heuristics explore the neighborhood of integer solutions by solving sub-MILPs, while maintaining the logic inherent to the upper and lower bounds of the B&C algorithm. Another interesting feature is the possibility of using multiple CPU threads to solve LP problems, and search the B&C tree. The solution of Boolean Quadratic Problems (BQP) using MILP reformulations, by direct solution with CPLEX, GUROBI, and XPRESS as a Mixed Integer Quadratic Programming (MIQP) problem, and using callbacks with CPLEX will be also covered. The convexification of BQP problems and the consequent solution by CPLEX will be briefly addressed. The results of computational experiences and several examples that evaluate the performance of these methods will be presented and discussed.
Ricardo M. Lima is a Marie Curie Fellow at the National Laboratory of Energy and Geology (LNEG) in Lisbon, Portugal. He received the Licentiate degree in 1999, and the Ph.D. degree in 2006, both in Chemical Engineering from the Faculty of Engineering, University of Porto, Portugal. He has worked with Ignacio E. Grossmann as a post-doc fellow in the Department of Chemical Engineering at the Carnegie Mellon University (CMU), PA, USA in 2006-2008, and has continued as a Researcher at the CMU in 2008-2011, with a joint position as Invited Researcher in PPG Industries in 2008-2011. He joined LNEG in 2011. His main research interests include mathematical programming, robust optimization, applied optimization, energy systems, and the design and scheduling/planning of industrial processes.