arXiv subjects

Marcia Fampa

Publications and source records attributed to Marcia Fampa.

34 records · Page 2Linked to original sources

arXiv2023

Computing D-Optimal solutions for huge-scale linear and quadratic response-surface models

We consider algorithmic approaches to the D-optimality problem for cases where the input design matrix is large and highly structured, in particular implicitly specified as a full quadratic or linear response-surface model in several levels of several factors. Using row generation techniques of mathematical optimization, in the context of discrete local-search and continuous relaxation aimed at branch-and-bound solution, we are able to design practical algorithms.

math.OC↗

arXiv2023

Branch-and-bound for D-Optimality with fast local search and variable-bound tightening

We apply a branch-and-bound (B\&B) algorithm to the D-optimality problem based on a convex mixed-integer nonlinear formulation. We discuss possible methodologies to accelerate the convergence of the B\&B algorithm, by combining the use of different upper bounds, variable-bound tightening inequalities, and local-search procedures. Different methodologies to compute the determinant of a matrix after a rank-one update are investigated to accelerate the local-searches. We discuss our findings through numerical experiments with randomly generated test problem.

math.OC↗

arXiv2023

Generalized Scaling for the Constrained Maximum-Entropy Sampling Problem

The best techniques for the constrained maximum-entropy sampling problem, a discrete-optimization problem arising in the design of experiments, are via a variety of concave continuous relaxations of the objective function. A standard bound-enhancement technique in this context is scaling. We extend this technique to generalized scaling, we give mathematical results aimed at supporting algorithmic methods for computing optimal generalized scalings, and we give computational results demonstrating the usefulness of generalized scaling on benchmark problem instances.

math.OC↗

arXiv2022

D-optimal Data Fusion: Exact and Approximation Algorithms

We study the D-optimal Data Fusion (DDF) problem, which aims to select new data points, given an existing Fisher information matrix, so as to maximize the logarithm of the determinant of the overall Fisher information matrix. We show that the DDF problem is NP-hard and has no constant-factor polynomial-time approximation algorithm unless P $=$ NP. Therefore, to solve the DDF problem effectively, we propose two convex integer-programming formulations and investigate their corresponding complementary and Lagrangian-dual problems. We also develop scalable randomized-sampling and local-search algorithms with provable performance guarantees. Leveraging the concavity of the objective functions in the two proposed formulations, we design an exact algorithm, aimed at solving the DDF problem to optimality. We further derive a family of submodular valid inequalities and optimality cuts, which can significantly enhance the algorithm performance. Finally, we test our algorithms using real-world data on the new phasor-measurement-units placement problem for modern power grids, considering the existing conventional sensors. Our numerical study demonstrates the efficiency of our exact algorithm and the scalability and high-quality outputs of our approximation algorithms.

math.OC↗

arXiv2022

On computing with some convex relaxations for the maximum-entropy sampling problem

Based on a factorization of an input covariance matrix, we define a mild generalization of an upper bound of Nikolov (2015) and Li and Xie (2020) for the NP-Hard constrained maximum-entropy sampling problem (CMESP). We demonstrate that this factorization bound is invariant under scaling and also independent of the particular factorization chosen. We give a variable-fixing methodology that could be used in a branch-and-bound scheme based on the factorization bound for exact solution of CMESP. We report on successful experiments with a commercial nonlinear-programming solver. We further demonstrate that the known "mixing" technique can be successfully used to combine the factorization bound with the factorization bound of the complementary CMESP, and also with the "linx bound" of Anstreicher (2020).

math.OC↗

arXiv2021

Experimental analysis of local searches for sparse reflexive generalized inverses

The well-known M-P (Moore-Penrose) pseudoinverse is used in several linear-algebra applications; for example, to compute least-squares solutions of inconsistent systems of linear equations. Irrespective of whether a given matrix is sparse, its M-P pseudoinverse can be completely dense, potentially leading to high computational burden and numerical difficulties, especially when we are dealing with high-dimensional matrices. The M-P pseudoinverse is uniquely characterized by four properties, but not all of them need to be satisfied for some applications. In this context, Fampa and Lee (Oper. Res. Letters, 46:605--610, 2018) and Xu, Fampa, Lee and Ponte (SIAM J. on Optimization, to appear) propose local-search procedures to construct sparse block-structured generalized inverses that satisfy only some of the M-P properties. (Vector) 1-norm minimization is used to induce sparsity and to keep the magnitude of the entries under control, and theoretical results limit the distance between the 1-norm of the solution of the local searches and the minimum 1-norm of generalized inverses with corresponding properties. We have implemented several local-search procedures based on results presented in these two papers and make here an experimental analysis of them, considering their application to randomly generated matrices of varied dimensions, ranks, and densities. Further, we carried out a case study on a real-world data set.

math.OC↗

arXiv2021

Masking Anstreicher's linx bound for improved entropy bounds

The maximum-entropy sampling problem is the NP-hard problem of maximizing the (log) determinant of an order-$s$ principle submatrix of a given order $n$ covariance matrix $C$. Exact algorithms are based on a branch-and-bound framework. The problem has wide applicability in spatial statistics, and in particular in environmental monitoring. Probably the best upper bound for the maximum, empirically, is Anstreicher's scaled ``linx'' bound (see [K.M. Anstreicher. Efficient solution of maximum-entropy sampling problems. Oper. Res., 68(6):1826--1835, 2020]). An earlier methodology for potentially improving any upper-bounding method is by masking; i.e. applying the bounding method to $C\circ M$, where $M$ is any correlation matrix. We establish that the linx bound can be improved via masking by an amount that is at least linear in $n$, even when optimal scaling parameters are employed.

math.OC↗

arXiv2021

Arc-flow approach for single batch-processing machine scheduling

We address the problem of scheduling jobs with non-identical sizes and distinct processing times on a single batch processing machine, aiming at minimizing the makespan. The extensive literature on this NP-hard problem mostly focuses on heuristics. Using an arc flow-based optimization approach, we construct an ingenious formulation that represents it as a problem of determining flows in graphs. The size of the formulation increases with the number of distinct sizes and processing times among the jobs, but it does not increase with the number of jobs, which makes it very effective to solve large instances to optimality, especially when multiple jobs have equal size and processing time. We compare our model to other models from the literature showing its clear superiority on benchmark instances, and proving optimality of random instances with up to 100 million jobs.

math.OC↗

arXiv2020

Trading off 1-norm and sparsity against rank for linear models using mathematical optimization: 1-norm minimizing partially reflexive ah-symmetric generalized inverses

The M-P (Moore-Penrose) pseudoinverse has as a key application the computation of least-squares solutions of inconsistent systems of linear equations. Irrespective of whether a given input matrix is sparse, its M-P pseudoinverse can be dense, potentially leading to high computational burden, especially when we are dealing with high-dimensional matrices. The M-P pseudoinverse is uniquely characterized by four properties, but only two of them need to be satisfied for the computation of least-squares solutions. Fampa and Lee (2018) and Xu, Fampa, Lee, and Ponte (2019) propose local-search procedures to construct sparse block-structured generalized inverses that satisfy the two key M-P properties, plus one more (the so-called reflexive property). That additional M-P property is equivalent to imposing a minimum-rank condition on the generalized inverse. (Vector) 1-norm minimization is used to induce sparsity and, importantly, to keep the magnitudes of entries under control for the generalized-inverses constructed. Here, we investigate the trade-off between low 1-norm and low rank for generalized inverses that can be used in the computation of least-squares solutions. We propose several algorithmic approaches that start from a $1$-norm minimizing generalized inverse that satisfies the two key M-P properties, and gradually decrease its rank, by iteratively imposing the reflexive property. The algorithms iterate until the generalized inverse has the least possible rank. During the iterations, we produce intermediate solutions, trading off low 1-norm (and typically high sparsity) against low rank.

math.OC↗

arXiv2020

Approximate 1-norm minimization and minimum-rank structured sparsity for various generalized inverses via local search

Fundamental in matrix algebra and its applications, a \emph{generalized inverse} of a real matrix $A$ is a matrix $H$ that satisfies the Moore-Penrose (M-P) property $AHA=A$. If $H$ also satisfies the additional useful M-P property, $HAH=H$, it is called a \emph{reflexive generalized inverse}. Reflexivity is equivalent to minimum rank, so we are particularly interested in reflexive generalized inverses. We consider aspects of symmetry related to the calculation of a \emph{sparse} reflexive generalized inverse of $A$. As is common, and following Lee and Fampa (2018) for calculating sparse generalized inverses, we use (vector) 1-norm minimization for inducing sparsity and for keeping the magnitude of entries under control. When $A$ is symmetric, we may naturally desire a symmetric $H$; while generally such a restriction on $H$ may not lead to a 1-norm minimizing reflexive generalized inverse. We investigate a block construction method to produce a symmetric reflexive generalized inverse that is structured and has guaranteed sparsity. We provide a theoretically-efficient and practical local-search algorithm to block-construct an approximate 1-norm minimizing symmetric reflexive generalized inverse. Another aspect of symmetry that we consider relates to another M-P property: $H$ is \emph{ah-symmetric} if $AH$ is symmetric. The ah-symmetry property is the key one for solving least-squares problems using $H$. Here we do not assume that $A$ is symmetric, and we do not impose symmetry on $H$. We investigate a column block construction method to produce an ah-symmetric reflexive generalized inverse that is structured and has guaranteed sparsity. We provide a theoretically-efficient and practical local-search algorithm to column block construct an approximate 1-norm minimizing ah-symmetric reflexive generalized inverse.

math.OC↗

arXiv2020

1-norm minimization and minimum-rank structured sparsity for symmetric and ah-symmetric generalized inverses: rank one and two

Generalized inverses are important in statistics and other areas of applied matrix algebra. A \emph{generalized inverse} of a real matrix $A$ is a matrix $H$ that satisfies the Moore-Penrose (M-P) property $AHA=A$. If $H$ also satisfies the M-P property $HAH=H$, then it is called \emph{reflexive}. Reflexivity of a generalized inverse is equivalent to minimum rank, a highly desirable property. We consider aspects of symmetry related to the calculation of various \emph{sparse} reflexive generalized inverses of $A$. As is common, we use (vector) 1-norm minimization for both inducing sparsity and for keeping the magnitude of entries under control. When $A$ is symmetric, a symmetric $H$ is highly desirable, but generally such a restriction on $H$ will not lead to a 1-norm minimizing reflexive generalized inverse. We investigate a block construction method to produce a symmetric reflexive generalized inverse that is structured and has guaranteed sparsity. Letting the rank of $A$ be $r$, we establish that the 1-norm minimizing generalized inverse of this type is a 1-norm minimizing symmetric generalized inverse when (i) $r=1$ and when (ii) $r=2$ and $A$ is nonnegative. Another aspect of symmetry that we consider relates to another M-P property: $H$ is \emph{ah-symmetric} if $AH$ is symmetric. The ah-symmetry property is sufficient for a generalized inverse to be used to solve the least-squares problem $\min\{\|Ax-b\|_2:~x\in\mathbb{R}^n\}$ using $H$, via $x:=Hb$. We investigate a column block construction method to produce an ah-symmetric reflexive generalized inverse that is structured and has guaranteed sparsity. We establish that the 1-norm minimizing ah-symmetric generalized inverse of this type is a 1-norm minimizing ah-symmetric generalized inverse when (i) $r=1$ and when (ii) $r=2$ and $A$ satisfies a technical condition.

math.OC↗

arXiv2020

Mixing convex-optimization bounds for maximum-entropy sampling

The maximum-entropy sampling problem is a fundamental and challenging combinatorial-optimization problem, with application in spatial statistics. It asks to find a maximum-determinant order-$s$ principal submatrix of an order-$n$ covariance matrix. Exact solution methods for this NP-hard problem are based on a branch-and-bound framework. Many of the known upper bounds for the optimal value are based on convex optimization. We present a methodology for "mixing" these bounds to achieve better bounds.

math.OC↗

arXiv2019

Parametric Convex Quadratic Relaxation of the Quadratic Knapsac Problem

We consider a parametric convex quadratic programming, CQP, relaxation for the quadratic knapsack problem, QKP. This relaxation maintains partial quadratic information from the original QKP by perturbing the objective function to obtain a concave quadratic term. The nonconcave part generated by the perturbation is then linearized by a standard approach that lifts the problem to the matrix space. We present a primal-dual interior point method to optimize the perturbation of the quadratic function, in a search for the tightest upper bound for the \QKPp. We prove that the same perturbation approach, when applied in the context of semidefinite programming, SDP, relaxations of the QKP, cannot improve the upper bound given by the corresponding linear \SDP relaxation. The result also applies to more general integer quadratic problems. Finally, we propose new valid inequalities on the lifted matrix variable, derived from cover and knapsack inequalities for the QKP, and present the separation problems to generate cuts for the current solution of the CQP relaxation. Our best bounds are obtained from alternating between optimizing the parametric quadratic relaxation over the perturbation and adding cutting planes generated by the valid inequalities proposed.

math.OC↗

arXiv2018

On Sparse Reflexive Generalized Inverses

We study sparse generalized inverses $H$ of a rank-$r$ real matrix $A$. We give a construction for reflexive generalized inverses having at most $r^2$ nonzeros. For $r=1$ and for $r=2$ with $A$ nonnegative, we demonstrate how to minimize the (vector) 1-norm over reflexive generalized inverses. For general $r$, we efficiently find reflexive generalized inverses with 1-norm within approximately a factor of $r^2$ of the minimum 1-norm generalized inverse.

math.OC↗

arXiv2018

Efficient treatment of bilinear forms in global optimization

We efficiently treat bilinear forms in the context of global optimization, by applying McCormick convexification and by extending an approach of Saxena, Bonami and Lee for symmetric quadratic forms to bilinear forms. A key application of our work is in treating "structural convexity" in a symmetric quadratic form.

math.OC↗

arXiv2016

Sparse pseudoinverses via LP and SDP relaxations of Moore-Penrose

Pseudoinverses are ubiquitous tools for handling over- and under-determined systems of equations. For computational efficiency, sparse pseudoinverses are desirable. Recently, sparse left and right pseudoinverses were introduced, using 1-norm minimization and linear programming. We introduce several new sparse pseudoinverses by developing linear and semi-definite programming relaxations of the well-known Moore-Penrose properties.

math.NA↗