Time: Thursday, 05/Sept/2024: 2:00pm - 3:30pm Session Chair: Tristan Gally |
Location: Theresianum 2607 Room Location at NavigaTUM |
Uno, a next-gen solver for unifying nonlinearly constrained nonconvex optimization
1Argonne National Laboratory, USA & ZIB, Germany; 2Argonne National Laboratory, USA
Derivative-based iterative methods for nonlinearly constrained nonconvex optimization usually share common algorithmic components, such as strategies for computing a descent direction and mechanisms that promote global convergence. Based on this observation, we introduce an abstract framework with four common ingredients that describes most derivative-based iterative methods and unifies their workflows. We then present Uno, a modular C++ solver that implements our abstract framework and allows the automatic generation of various strategy combinations with no programming effort from the user. Uno is meant to
(1) organize mathematical optimization strategies into a coherent hierarchy;
(2) offer a wide range of efficient and robust methods that can be compared for a given instance;
(3) reduce the cost of development and maintenance of multiple optimization solvers; and
(4) enable researchers to experiment with novel optimization strategies while leveraging established subproblem solvers and interfaces to modeling languages.
We demonstrate that Uno is highly competitive against state-of-the-art solvers such as filterSQP, IPOPT, SNOPT, MINOS, LANCELOT, LOQO and CONOPT on a set of 429 problems from the CUTEst collection.
Uno is available as open-source software under the MIT license at https://github.com/cvanaret/Uno.
BOBILib: Bilevel Optimization (Benchmark) Instance Library
1Trier University, Germany; 2Quantagonia; 3ESSEC Business School of Paris, Cergy-Pontoise, France; 4Department of Industrial and System Engineering, Lehigh University, Bethlehem, PA, USA
The field of computational mixed-integer linear bilevel optimization made significant advances over the last 15 years. Many new problem classes have been tackled and a significant number of algorithmic techniques have been introduced. Consequently, we can solve new types of or simply larger problems compared to what was possible in the 2000s. However, what was missing up to now, was a well-curated and publicly available test-set of instances that the community can use to develop, test, and compare novel methods. In this talk, we present the Bilevel Optimization Benchmark Instance Library (BOBILib), which aims for filling the above mentioned gap. BOBILib contains over 2000 instances ranging from general mixed-integer linear bilevel problems to specific interdiction problems. In this talk, we will discuss the structure of the library, the type of instances involved, and the data format in which the problems are given.
Solving MINLPs to Global Optimality with FICO Xpress
FICO
This talk discusses the global optimization capability within FICO Xpress Solver, which allows to solve general mixed-integer nonlinear problems to global optimality. We will explain the internal workings of the solver, its features, and recent performance improvements.