The CPLEX team will be in Seattle, WA, for the 2019 INFORMS Annual Meeting. We have a fully packed agenda for this event:
– Saturday, Oct. 19, 1-3:30pm, Pre-conference workshop.
– Come meet us during the whole event at booth #25
– Oct. 21, 11:00 AM, MB33, Recent Progress In CPLEX Benders Decomposition
– Oct. 21, 4:30 PM, MD33, What’s new in CPLEX
– Oct. 22, 4:35 PM, TE35, Generalized Surrogate Duality for Mixed-Integer Nonlinear Programs.
– Oct. 23, 1:30 PM, WC40, Penalty Variables and Multiobjective Optimization
– Oct. 23, 3:20 PM, WD44, Applying a Classifier to Solve Mixed Integer Quadratic Problems in CPLEX
Abstracts of the various presentations:
Recent progress in IBM CPLEX Optimization Studio and Decision Optimization Center
Speakers: Dr Pierre Bonami, Dr Filippo Focacci, Dr Daniel Junglas, Dr Roland Wunderling
The first part of the workshop will be focused on CPLEX Optimization Studio 12.10. This new version brings performance improvements and additional features. In particular, our MIQP solver uses a Machine Learning model to decide whether to linearise the objective or not, for better decisions about this crucial step, and the implementation of automatic Benders decomposition also offers significantly improved performance. The new callback framework introduced in the previous version was enriched with a branching callback that allows users to guide CPLEX in its exploration of the tree. We will present these improvements, and more.
DOC is an application development platform delivering optimization applications. It builds on various opensource technologies which allows it to be an efficient delivery mechanism both onprem and cloud through docker containers. To facilitate end-user access to data and solutions through web-visualization an Angular.js-based web-GUI framework is provided with out-of-the-box charts to ease GUI development. In the second part of this workshop it will be shown how a complete application can be delivered in around thirty minutes it the user has the optimisation model ready and has decided about the architecture (client/server) and what visual components should be used in the final application.
October 21, 2019, 11:00 AM, MB33
Recent Progress In CPLEX Benders Decomposition
Dr Andrea Tramontani
In this talk we present the Benders decomposition branch-and-cut that is implemented in CPLEX for Mixed Integer Linear Programming (MILP). We illustrate the main algorithmic components behind our implementation and discuss the latest improvements that are currently work in progress. Finally, we present an extensive computational analysis on some classes of decomposable MILP problems, to assess the performance of Benders branch-and-cut in comparison with the default branch-and-cut of CPLEX. The results show that some models that are out of reach for a “standard” branch-and-cut can instead be solved by Benders decomposition.
October 21, 2019, 4:30 PM, MD33
What’s new in CPLEX
Dr Roland Wunderling
We will present the performance improvements realized in the latest release of CPLEX and describe the ideas that helped achieve them.
October 22, 2019, 4:35 PM, TE35
Generalized Surrogate Duality for Mixed-Integer Nonlinear Programs.
Prof. Andrea Lodi
The most important ingredient for solving mixed-integer nonlinear programs (MINLPs) to global optimality with spatial branch-and-bound is a tight, computationally tractable relaxation. Due to both theoretical and practical considerations, these relaxations are usually required to be convex. Nonetheless, current optimization solvers can often handle a moderate presence of non-convexities efficiently, which opens the door for the use of potentially tighter non-convex relaxations. One way to construct such a relaxation is by aggregating nonlinear constraints of the problem into a single non-convex constraint. The resulting relaxation is called surrogate relaxation, which is similar to the well-known Lagrangian relaxation but provides stronger dual bounds in general. In this talk, we make use of the surrogate dual in a non-convex setting and show the benefits and challenges such relaxation can have in general MINLPs. Additionally, we present a generalization of the surrogate relaxation that allows for multiple aggregations of non-convex constraints. Based on a known Benders-type approach, we present an algorithm that can solve our generalized surrogate relaxation, along with several computational enhancements for improving its practical performance. Finally, we conduct extensive computational experiments on instances of the MINLPLib using the global solver SCIP.
October 23, 2019, 1:30 PM, WC40
Penalty Variables and Multiobjective Optimization
With the CPLEX 12.9 release it is possible to define multiple objectives for a single problem. In this talk we will introduce a simplified definition of a penalty variable, and present a method for taking advantage of penalty variables with a multiobjective formulation.
Finally, we will compare the performance of this method against default CPLEX.
October 23, 2019, 3:20 PM, WD44
Applying a Classifier to Solve Mixed Integer Quadratic Problems in CPLEX
Dr Pierre Bonami
Within state-of-the-art optimization solvers such as IBM-CPLEX the ability to solve both convex and nonconvex Mixed-Integer Quadratic Programming (MIQP) problems to proven optimality goes back few years, but still presents unclear aspects. We are interested in understanding whether for solving an MIQP problem it is favorable to linearize its quadratic part or not. Our approach employs Machine Learning techniques to learn a classifier that predicts, for a given MIQP instance, the most suitable resolution method within IBM-CPLEX’s algorithmic framework.