## 90C26 Nonconvex programming, global optimization

### Refine

#### Document Type

- ZIB-Report (17)
- Doctoral Thesis (2)

#### Keywords

- Pooling Problem (4)
- nonconvex (3)
- Cutting Planes (2)
- Large Neighborhood Search (2)
- MINLP (2)
- Nonconvexity (2)
- Primal Heuristic (2)
- Quadratic Programming (2)
- Relaxation (2)
- Relaxations (2)

#### Institute

The amazing success of computational mathematical optimization over the last decades has been driven more by insights into mathematical structures than by the advance of computing technology. In this vein, we address applications, where nonconvexity in the model poses principal difficulties.
This paper summarizes the dissertation of Jonas Schweiger for the occasion of the GOR dissertation award 2018. We focus on the work on non-convex quadratic programs and show how problem specific structure can be used to obtain tight relaxations and speed up Branch&Bound methods. Both a classic general QP and the Pooling Problem as an important practical application serve as showcases.

We investigate new convex relaxations for the pooling problem, a classic nonconvex production planning problem in which products are mixed in intermediate pools in order to meet quality targets at their destinations. In this technical report, we characterize the extreme points of the convex hull of our non-convex set, and show that they are not finite, i.e., the convex hull is not polyhedral. This analysis was used to derive valid nonlinear convex inequalities and show that, for a specific case, they characterize the convex hull of our set. The new valid inequalities and computational results are presented in ZIB Report 18-12.

We investigate new convex relaxations for the pooling problem, a classic nonconvex production planning problem in which input materials are mixed in intermediate pools, with the outputs of these pools further mixed to make output products meeting given attribute percentage requirements. Our relaxations are derived by considering a set which arises from the formulation by considering a single product, a single attibute, and a single pool. The convex hull of the resulting nonconvex set is not polyhedral. We derive valid linear and convex nonlinear inequalities for the convex hull, and demonstrate that different subsets of these inequalities define the convex hull of the nonconvex set in three cases determined by the parameters of the set. Computational results on literature instances and newly created larger test instances demonstrate that the inequalities can significantly strengthen the convex relaxation of the pq-formulation of the pooling problem, which is the relaxation known to have the strongest bound.

Exploiting structure in non-convex quadratic optimization and gas network planning under uncertainty
(2017)

The amazing success of computational mathematical optimization over
the last decades has been driven more by insights into mathematical
structures than by the advance of computing technology. In this vein,
we address applications, where nonconvexity in the model and
uncertainty in the data pose principal difficulties.
The first part of the thesis deals with non-convex quadratic programs.
Branch&Bound methods for this problem class depend on tight
relaxations. We contribute in several ways: First, we establish a new
way to handle missing linearization variables in the well-known
Reformulation-Linearization-Technique (RLT). This is implemented
into the commercial software CPLEX. Second, we study the optimization
of a quadratic objective over the standard simplex or a knapsack
constraint. These basic structures appear as part of many complex
models. Exploiting connections to the maximum clique problem and RLT,
we derive new valid inequalities. Using exact and heuristic separation
methods, we demonstrate the impact of the new inequalities on the
relaxation and the global optimization of these problems. Third, we
strengthen the state-of-the-art relaxation for the pooling problem, a
well-known non-convex quadratic problem, which is, for example,
relevant in the petrochemical industry. We propose a novel relaxation
that captures the essential non-convex structure of the problem but is
small enough for an in-depth study. We provide a complete inner
description in terms of the extreme points as well as an outer
description in terms of inequalities defining its convex hull (which
is not a polyhedron). We show that the resulting valid convex
inequalities significantly strengthen the standard relaxation of the
pooling problem.
The second part of this thesis focuses on a common challenge in real
world applications, namely, the uncertainty entailed in the input
data.
We study the extension of a gas transport network, e.g., from our
project partner Open Grid Europe GmbH.
For a single scenario this maps to a challenging non-convex MINLP.
As the future transport patterns are highly uncertain, we propose a
robust model to best prepare the network operator for an array of
scenarios.
We develop a custom decomposition approach that makes use of the
hierarchical structure of network extensions and the loose coupling
between the scenarios.
The algorithm used the single-scenario problem as black-box subproblem
allowing the generalization of our approach to problems with the same
structure.
The scenario-expanded version of this problem is out of reach for
today's general-purpose MINLP solvers.
Yet our approach provides primal and dual bounds for instances with up
to 256 scenarios and solves many of them to optimality.
Extensive computational studies show the impact of our work.

The SCIP Optimization Suite is a powerful collection of optimization software that consists of the branch-cut-and-price framework and mixed-integer programming solver SCIP, the linear programming solver SoPlex, the modeling language Zimpl, the parallelization framework UG, and the generic branch-cut-and-price solver GCG. Additionally, it features the extensions SCIP-Jack for solving Steiner tree problems, PolySCIP for solving multi-objective problems, and SCIP-SDP for solving mixed-integer semidefinite programs. The SCIP Optimization Suite has been continuously developed and has now reached version 4.0. The goal of this report is to present the recent changes to the collection. We not only describe the theoretical basis, but focus on implementation aspects and their computational consequences.

This paper describes the extensions that were added to the constraint integer programming framework SCIP in order to enable it to solve convex and nonconvex mixed-integer nonlinear programs (MINLPs) to global optimality.
SCIP implements a spatial branch-and-bound algorithm based on a linear outer-approximation, which is computed by convex over- and underestimation of nonconvex functions. An expression graph representation of nonlinear constraints allows for bound tightening, structure analysis, and reformulation. Primal heuristics are employed throughout the solving process to find feasible solutions early. We provide insights into the performance impact of individual MINLP solver components via a detailed computational study over a large and heterogeneous test set.

In the literature for mixed integer programming, heuristic algorithms (particularly primal heuristics) are often considered as stand-alone procedures; in that context, heuristics are treated as an alternative to solving a problem to proven optimality. This conceals the fact that heuristic algorithms are a fundamental component of state-of-the-art global solvers for mixed integer linear programming (MIP) and mixed integer nonlinear programming (MINLP).
In the present thesis, we focus on this latter aspect; we study heuristic algorithms that are tightly integrated within global MINLP solvers and analyze their impact on the overall solution process. Our contributions comprise generalizations of primal heuristics for MIP towards MINLP as well as novel ideas for MINLP primal heuristics and for heuristic algorithms to take branching decisions and to collect global information in MIP. These are:
- Shift-and-Propagate, a novel propagation heuristic for MIP that does not require the solution to an LP relaxation,
- a generic way to generalize large neighborhood search (LNS) heuristics from MIP to MINLP,
- an Objective Feasibility Pump heuristic for nonconvex MINLP that uses second-order information and a dynamic selection of rounding procedures,
- RENS, an LNS start heuristic for MINLP that optimizes over the set of feasible roundings of an LP solution,
- Undercover, an LNS start heuristic for MINLP that solves a largest sub-MIP of a given MINLP,
- Rapid Learning, a heuristic algorithm to generate globally valid conflict constraints for MIPs,
- Cloud Branching, a heuristic algorithm that exploits dual degeneracy to reduce the number of candidates for branching variable selection.
Additionally, we propose a new performance measure, the primal integral, that captures the benefits of primal heuristics better than traditional methods. In our computational study, we compare the performance of the MIP and MINLP solver SCIP with and without primal heuristics on six test sets with altogether 983 instances from academic and industrial sources, including our project partners ForNe, SAP, and Siemens. We observe that heuristics improve the solver performance regarding all measures that we used - by different orders of magnitude. We further see that the harder a problem is to solve to global optimality, the more important the deployment of primal heuristics becomes.
The algorithms presented in this thesis are available in source code as part of the solver SCIP, of which the author has been a main developer for the last years. Methods described in this thesis have also been re-implemented within several commercial and noncommercial MIP and MINLP software packages, including Bonmin, CBC, Cplex, Gams, Sulum, and Xpress.

Primal heuristics are an important component of state-of-the-art codes for mixed integer nonlinear programming (MINLP). In this article we give a compact overview of primal heuristics for MINLP that have been suggested in the literature of recent years. We sketch the fundamental concepts of different classes of heuristics and discuss specific implementations. A brief computational experiment shows that primal heuristics play a key role in achieving feasibility and finding good primal bounds within a global MINLP solver.

Undercover Branching
(2013)

In this paper, we present a new branching strategy for nonconvex MINLP that aims at driving the created subproblems towards linearity. It exploits the structure of a minimum cover of an MINLP, a smallest set of variables that, when fixed, render the remaining system linear: whenever possible, branching candidates in the cover are preferred.
Unlike most branching strategies for MINLP, Undercover branching is not an extension of an existing MIP branching rule. It explicitly regards the nonlinearity of the problem while branching on integer variables with a fractional relaxation solution. Undercover branching can be naturally combined with any variable-based branching rule.
We present computational results on a test set of general MINLPs from MINLPLib, using the new strategy in combination with reliability branching and pseudocost branching. The computational cost of Undercover branching itself proves negligible. While it turns out that it can influence the variable selection only on a smaller set of instances, for those that are affected, significant improvements in performance are achieved.

We provide a computational study of the performance of a state-of-the-art solver for nonconvex mixed-integer quadratically constrained programs (MIQCPs). Since successful general-purpose solvers for large problem classes necessarily comprise a variety of algorithmic techniques, we focus especially on the impact of the individual solver components. The solver SCIP used for the experiments implements a branch-and-cut algorithm based on a linear relaxation to solve MIQCPs to global optimality. Our analysis is based on a set of 86 publicly available test instances.