arXiv subjects

Kathrin Klamroth

Publications and source records attributed to Kathrin Klamroth.

30 records · Page 2Linked to original sources

arXiv2022

Biobjective Optimization Problems on Matroids with Binary Costs

Like most multiobjective combinatorial optimization problems, biobjective optimization problems on matroids are in general intractable and their corresponding decision problems are in general NP-hard. In this paper, we consider biobjective optimization problems on matroids where one of the objective functions is restricted to binary cost coefficients. We show that in this case the problem has a connected efficient set with respect to a natural definition of a neighborhood structure and hence, can be solved efficiently using a neighborhood search approach. This is, to the best of our knowledge, the first non-trivial problem on matroids where connectedness of the efficient set can be established. The theoretical results are validated by numerical experiments with biobjective minimum spanning tree problems (graphic matroids) and with biobjective knapsack problems with a cardinality constraint (uniform matroids). In the context of the minimum spanning tree problem, coloring all edges with cost 0 green and all edges with cost 1 red leads to an equivalent problem where we want to simultaneously minimize one general objective and the number of red edges (which defines the second objective) in a Pareto sense.

cs.DS↗

arXiv2021

Multi-objective Matroid Optimization with Ordinal Weights

Bi-objective optimization problems on matroids are in general intractable and their corresponding decision problems are in general NP-hard. However, if one of the objective functions is restricted to binary cost coefficients the problem becomes efficiently solvable by an exhaustive swap algorithm. Binary cost coefficients often represent two categories and are thus a special case of ordinal coefficients that are in general non-additive. In this paper we consider ordinal objective functions with more than two categories in the context of matroid optimization. We introduce several problem variants that can be distinguished w.r.t. their respective optimization goals, analyze their interrelations, and derive a polynomial time solution method that is based on the repeated solution of matroid intersection problems. Numerical tests on minimum spanning tree problems and on partition matroids confirm the efficiency of the approach.

math.OC↗

arXiv2020

Efficient and Sparse Neural Networks by Pruning Weights in a Multiobjective Learning Approach

Overparameterization and overfitting are common concerns when designing and training deep neural networks, that are often counteracted by pruning and regularization strategies. However, these strategies remain secondary to most learning approaches and suffer from time and computational intensive procedures. We suggest a multiobjective perspective on the training of neural networks by treating its prediction accuracy and the network complexity as two individual objective functions in a biobjective optimization problem. As a showcase example, we use the cross entropy as a measure of the prediction accuracy while adopting an l1-penalty function to assess the total cost (or complexity) of the network parameters. The latter is combined with an intra-training pruning approach that reinforces complexity reduction and requires only marginal extra computational cost. From the perspective of multiobjective optimization, this is a truly large-scale optimization problem. We compare two different optimization paradigms: On the one hand, we adopt a scalarization-based approach that transforms the biobjective problem into a series of weighted-sum scalarizations. On the other hand we implement stochastic multi-gradient descent algorithms that generate a single Pareto optimal solution without requiring or using preference information. In the first case, favorable knee solutions are identified by repeated training runs with adaptively selected scalarization parameters. Preliminary numerical results on exemplary convolutional neural networks confirm that large reductions in the complexity of neural networks with neglibile loss of accuracy are possible.

cs.LG↗

arXiv2020

Tracing locally Pareto optimal points by numerical integration

We suggest a novel approach for the efficient and reliable approximation of the Pareto front of sufficiently smooth unconstrained bi-criteria optimization problems. Optimality conditions formulated for weighted sum scalarizations of the problem yield a description of (parts of) the Pareto front as a parametric curve, parameterized by the scalarization parameter (i.e., the weight in the weighted sum scalarization). Its sensitivity w.r.t. parameter variations can be described by an ordinary differential equation (ODE). Starting from an arbitrary initial Pareto optimal solution, the Pareto front can then be traced by numerical integration. We provide an error analysis based on Lipschitz properties and suggest an explicit Runge-Kutta method for the numerical solution of the ODE. The method is validated on bi-criteria convex quadratic programming problems for which the exact solution is explicitly known, and numerically tested on complex bi-criteria shape optimization problems involving finite element discretizations of the state equation.

math.OC↗

arXiv2020

GivEn -- Shape Optimization for Gas Turbines in Volatile Energy Networks

This paper describes the project GivEn that develops a novel multicriteria optimization process for gas turbine blades and vanes using modern "adjoint" shape optimization algorithms. Given the many start and shut-down processes of gas power plants in volatile energy grids, besides optimizing gas turbine geometries for efficiency, the durability understood as minimization of the probability of failure is a design objective of increasing importance. We also describe the underlying coupling structure of the multiphysical simulations and use modern, gradient based multicriteria optimization procedures to enhance the exploration of Pareto-optimal solutions.

math.OC↗

arXiv2019

A simple and efficient dichotomic search algorithm for multi-objective mixed integer linear programs

We present a simple and at the same time fficient algorithm to compute all nondominated extreme points in the outcome set of multi-objective mixed integer linear programmes in any dimension. The method generalizes the well-known dichotomic scheme to compute the set of nondominated extreme points in the outcome set of a bi-objective programme based on the iterative solution of weighted sum scalarizations. It uses as a main routine a convex hull algorithm. The algorithm is illustrated with, and numerically tested on, instances of multi-objective assignment and knapsack problems. Experimental results confirm the computational efficiency of the approach. Finally, an implementation in incremental polynomial time with respect to the number of computed nondominated extreme points is possible, under the assumption that the lexicographic version of the problem can be solved in polynomial time.

math.OC↗

arXiv2019

Gradient Based Biobjective Shape Optimization to Improve Reliability and Cost of Ceramic Components

We consider the simultaneous optimization of the reliability and the cost of a ceramic component in a biobjective PDE constrained shape optimization problem. A probabilistic Weibull-type model is used to assess the probability of failure of the component under tensile load, while the cost is assumed to be proportional to the volume of the component. Two different gradient-based optimization methods are suggested and compared at 2D test cases. The numerical implementation is based on a first discretize then optimize strategy and benefits from efficient gradient computations using adjoint equations. The resulting approximations of the Pareto front nicely exhibit the trade-off between reliability and cost and give rise to innovative shapes that compromise between these conflicting objectives.

math.OC↗

arXiv2018

Introducing Multiobjective Complex Systems

This article focuses on the optimization of a complex system which is composed of several subsystems. On the one hand, these subsystems are subject to multiple objectives, local constraints as well as local variables, and they are associated with an own, subsystem-dependent decision maker. On the other hand, these subsystems are interconnected to each other by global variables or linking constraints. Due to these interdependencies, it is in general not possible to simply optimize each subsystem individually to improve the performance of the overall system. This article introduces a formal graph-based representation of such complex systems and generalizes the classical notions of feasibility and optimality to match this complex situation. Moreover, several algorithmic approaches are suggested and analyzed.

math.OC↗

arXiv2018

Adjoint Method to Calculate Shape Gradients of Failure Probabilaties for Turbomachinery Components

In the optimization of turbomachinery components, shape sensitivities for fluid dynamical objective functions have been used for a long time. As peak stress is not a differential func- tional of the shape, such highly efficient procedures so far have been missing for objective functionals that stem from mechan- ical integrity. This changes, if deterministic lifing criteria are replaced by probabilistic criteria, which have been introduced recently to the field of low cycle fatigue (LCF). Here we present a finite element (FEA) based first discretize, then adjoin approach to the calculation of shape gradients (sen- sitivities) for the failure probability with regard to probabilistic LCF and apply it to simple and complex geometries, as e.g. a blisk geometry. We review the computation of failure probabilities with a FEA postprocessor and sketch the computation of the relevant quantities for the adjoint method. We demonstrate high accuracy and computational efficiency of the adjoint method compared to finite difference schemes. We discuss implementation details for rotating components with cyclic boundary conditions. Finally, we shortly comment on future development steps and on poten- tial applications in multi criteria optimization.

math.NA↗

arXiv2015

A Box Decomposition Algorithm to Compute the Hypervolume Indicator

We propose a new approach to the computation of the hypervolume indicator, based on partitioning the dominated region into a set of axis-parallel hyperrectangles or boxes. We present a nonincremental algorithm and an incremental algorithm, which allows insertions of points, whose time complexities are $O(n^{\lfloor \frac{p-1}{2} \rfloor+1})$ and $O(n^{\lfloor \frac{p}{2} \rfloor+1})$, respectively. While the theoretical complexity of such a method is lower bounded by the complexity of the partition, which is, in the worst-case, larger than the best upper bound on the complexity of the hypervolume computation, we show that it is practically efficient. In particular, the nonincremental algorithm competes with the currently most practically efficient algorithms. Finally, we prove an enhanced upper bound of $O(n^{p-1})$ and a lower bound of $Ω(n^{\lfloor \frac{p}{2}\rfloor} \log n )$ for $p \geq 4$ on the worst-case complexity of the WFG algorithm.

cs.DM↗

arXiv2015

On the representation of the search region in multi-objective optimization

Given a finite set $N$ of feasible points of a multi-objective optimization (MOO) problem, the search region corresponds to the part of the objective space containing all the points that are not dominated by any point of $N$, i.e. the part of the objective space which may contain further nondominated points. In this paper, we consider a representation of the search region by a set of tight local upper bounds (in the minimization case) that can be derived from the points of $N$. Local upper bounds play an important role in methods for generating or approximating the nondominated set of an MOO problem, yet few works in the field of MOO address their efficient incremental determination. We relate this issue to the state of the art in computational geometry and provide several equivalent definitions of local upper bounds that are meaningful in MOO. We discuss the complexity of this representation in arbitrary dimension, which yields an improved upper bound on the number of solver calls in epsilon-constraint-like methods to generate the nondominated set of a discrete MOO problem. We analyze and enhance a first incremental approach which operates by eliminating redundancies among local upper bounds. We also study some properties of local upper bounds, especially concerning the issue of redundant local upper bounds, that give rise to a new incremental approach which avoids such redundancies. Finally, the complexities of the incremental approaches are compared from the theoretical and empirical points of view.

cs.DM↗

arXiv2014

A linear bound on the number of scalarizations needed to solve discrete tricriteria optimization problems

General multi-objective optimization problems are often solved by a sequence of parametric single objective problems, so-called scalarizations. If the set of nondominated points is finite, and if an appropriate scalarization is employed, the entire nondominated set can be generated in this way. In the bicriteria case it is well known that this can be realized by an adaptive approach which, given an appropriate initial search space, requires the solution of a number of subproblems which is at most two times the number of nondominated points. For higher dimensional problems, no linear methods were known up to now. We present a new procedure for finding the entire nondominated set of tricriteria optimization problems for which the number of scalarized subproblems to be solved is at most three times the number of nondominated points of the underlying problem. The approach includes an iterative update of the search space that, given a (sub-)set of nondominated points, describes the area in which additional nondominated points may be located. In particular, we show that the number of boxes, into which the search space is decomposed, depends linearly on the number of nondominated points.

math.OC↗