We gratefully acknowledge support from
the Simons Foundation and member institutions.

Optimization and Control

New submissions

[ total of 28 entries: 1-28 ]
[ showing up to 2000 entries per page: fewer | more ]

New submissions for Wed, 20 Oct 21

[1]  arXiv:2110.09569 [pdf, other]
Title: Constrained Discrete Black-Box Optimization using Mixed-Integer Programming
Comments: 10 pages, 4 figures, appendix with additional results
Subjects: Optimization and Control (math.OC)

Discrete black-box optimization problems are challenging for model-based optimization (MBO) algorithms, such as Bayesian optimization, due to the size of the search space and the need to satisfy combinatorial constraints. In particular, these methods require repeatedly solving a complex discrete global optimization problem in the inner loop, where popular heuristic inner-loop solvers introduce approximations and are difficult to adapt to combinatorial constraints. In response, we propose NN+MILP, a general discrete MBO framework using piecewise-linear neural networks as surrogate models and mixed-integer linear programming (MILP) to optimize the acquisition function. MILP provides optimality guarantees and a versatile declarative language for domain-specific constraints. We test our approach on a range of unconstrained and constrained problems, including DNA binding and the NAS-Bench-101 neural architecture search benchmark. NN+MILP surpasses or matches the performance of algorithms tailored to the domain at hand, with global optimization of the acquisition problem running in a few minutes using only standard software packages and hardware.

[2]  arXiv:2110.09678 [pdf, ps, other]
Title: Convergence Rate of Accelerated Average Consensus with Local Node Memory: Optimization and Analytic Solutions
Comments: 30 pages, 2 figures
Subjects: Optimization and Control (math.OC); Systems and Control (eess.SY)

Previous researches have shown that adding local memory can accelerate the consensus. It is natural to ask questions like what is the fastest rate achievable by the $M$-tap memory acceleration, and what are the corresponding control parameters. This paper introduces a set of effective and previously unused techniques to analyze the convergence rate of accelerated consensus with $M$-tap memory of local nodes and to design the control protocols. These effective techniques, including the Kharitonov stability theorem, the Routh stability criterion and the robust stability margin, have led to the following new results: 1) the direct link between the convergence rate and the control parameters; 2) explicit formulas of the optimal convergence rate and the corresponding optimal control parameters for $M \leq 2$ on a given graph; 3) the optimal worst-case convergence rate and the corresponding optimal control parameters for the memory $M \geq 1$ on a set of uncertain graphs. We show that the acceleration with the memory $M = 1$ provides the optimal convergence rate in the sense of worst-case performance. Several numerical examples are given to demonstrate the validity and performance of the theoretical results.

[3]  arXiv:2110.09694 [pdf, other]
Title: Robust-and-Cheap Framework for Network Resilience: A Novel Mixed-Integer Formulation and Solution Method
Subjects: Optimization and Control (math.OC); Combinatorics (math.CO)

Resilience and robustness are important properties in the reliability and attack-tolerance analysis of networks. In recent decades, various qualitative and heuristic-based quantitative approaches have made significant contributions in addressing network resilience and robustness. However, the lack of exact methods such as mixed-integer programming (MIP) models is sensible in the literature. In this paper, we contribute to the literature on the network resilience and robustness for targeted and random attacks and propose a MIP model considering graph-theoretical aspects of networks. The proposed MIP model consists of two stages where in the first stage the worst-case attack %leading to the most severe damage in the network is identified. Then, the second stage maximizes the network resilience under the worst-case attack by adding links considering a link addition financial budget. In addition, we propose a solution method that (i) provides a tight relaxation for the MIL formulation by relaxing some of the integrality restrictions, (ii) exploits the structure of the problem and reduces the second-stage problem to a less complex but equivalent problem, and (iii) identifies underlying knapsack constraints and generates lifted cover inequalities (LCI) cuts for identified knapsack constraints. We conclude numerical experiments for random networks and then extend our results to power system networks. Numerical experiments demonstrate the applicability and computational efficiency of the proposed robust-and-cheap framework for network resilience.

[4]  arXiv:2110.09738 [pdf, other]
Title: Faster Rates for the Frank-Wolfe Algorithm Using Jacobi Polynomials
Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG)

The Frank Wolfe algorithm (FW) is a popular projection-free alternative for solving large-scale constrained optimization problems. However, the FW algorithm suffers from a sublinear convergence rate when minimizing a smooth convex function over a compact convex set. Thus, exploring techniques that yield a faster convergence rate becomes crucial. A classic approach to obtain faster rates is to combine previous iterates to obtain the next iterate. In this work, we extend this approach to the FW setting and show that the optimal way to combine the past iterates is using a set of orthogonal Jacobi polynomials. We also a polynomial-based acceleration technique, referred to as Jacobi polynomial accelerated FW, which combines the current iterate with the past iterate using combing weights related to the Jacobi recursion. By carefully choosing parameters of the Jacobi polynomials, we obtain a faster sublinear convergence rate. We provide numerical experiments on real datasets to demonstrate the efficacy of the proposed algorithm.

[5]  arXiv:2110.09854 [pdf, ps, other]
Title: A New Extension of Chubanov's Method to Symmetric Cones
Comments: 47 pages
Subjects: Optimization and Control (math.OC); Numerical Analysis (math.NA)

We propose a new variant of Chubanov's method for solving the feasibility problem over the symmetric cone by extending Roos's method (2018) for the feasibility problem over the nonnegative orthant. The proposed method considers a feasibility problem associated with a norm induced by the maximum eigenvalue of an element and uses a rescaling focusing on the upper bound of the sum of eigenvalues of any feasible solution to the problem. Its computational bound is (i) equivalent to Roos's original method (2018) and superior to Louren\c{c}o et al.'s method (2019) when the symmetric cone is the nonnegative orthant, (ii) superior to Louren\c{c}o et al.'s method (2019) when the symmetric cone is a Cartesian product of second-order cones, and (iii) equivalent to Louren\c{c}o et al.'s method (2019) when the symmetric cone is the simple positive semidefinite cone, under the assumption that the costs of computing the spectral decomposition and the minimum eigenvalue are of the same order for any given symmetric matrix.
We also conduct numerical experiments that compare the performance of our method with existing methods by generating instance in three types: (i) strongly (but ill-conditioned) feasible instances, (ii) weakly feasible instances, and (iii) infeasible instances. For any of these instances, the proposed method is rather more efficient than the existing methods in terms of accuracy and execution time.

[6]  arXiv:2110.09901 [pdf, other]
Title: A FPTAS for the Subset Sum Problem with Real Numbers
Authors: Marius Costandin
Comments: arXiv admin note: text overlap with arXiv:2107.08482
Subjects: Optimization and Control (math.OC)

In this paper we study the subset sum problem with real numbers. Starting from the given problem, we formulate it as a quadratic maximization over a polytope. We provide a polynomial algorithm which maximizes the distance to a fixed point over a certain convex set. This convex set is obtained by intersecting the unit hypercube with two relevant half spaces. We show that in case the subset sum problem has a solution, our algorithm gives the correct maximum distance up to an arbitrary chosen precision, otherwise it will give a strictly less value. We show that if the subset sum problem has a solution then the maximum distance is expected to be a certain number, that can easily be computed beforehand. We than apply our algorithm to compute the distance and compare the result with what is expected and as such we can assert the feasibility of the subset sum problem.

[7]  arXiv:2110.09917 [pdf, ps, other]
Title: Planning for Package Deliveries in Risky Environments Over Multiple Epochs
Subjects: Optimization and Control (math.OC); Discrete Mathematics (cs.DM); Probability (math.PR)

We study a risk-aware robot planning problem where a dispatcher must construct a package delivery plan that maximizes the expected reward for a robot delivering packages across multiple epochs. Each package has an associated reward for delivery and a risk of failure. If the robot fails while delivering a package, no future packages can be delivered and the cost of replacing the robot is incurred. The package delivery plan takes place over the course of either a finite or an infinite number of epochs, denoted as the finite horizon problem and infinite horizon problem, respectively. The dispatcher has to weigh the risk and reward of delivering packages during any given epoch against the potential loss of any future epoch's reward. By using the ratio between a package's reward and its risk of failure, we prove an optimal, greedy solution to both the infinite and finite horizon problems. The finite horizon problem can be solved optimally in $O(K n\log n)$ time where $K$ is the number of epochs and $n$ is the number of packages. We show an isomorphism between the infinite horizon problem and Markov Decision Processes to prove an optimal $O(n)$ time algorithm for the infinite horizon problem.

Cross-lists for Wed, 20 Oct 21

[8]  arXiv:2110.09560 (cross-list from math.PR) [pdf, ps, other]
Title: On the optimality of the refraction--reflection strategy for Lévy processes
Authors: Kei Noba
Comments: 44 pages, 4 figures
Subjects: Probability (math.PR); Optimization and Control (math.OC)

In this paper, we study de Finetti's optimal dividend problem with capital injection under the assumption that the dividend strategies are absolutely continuous. In many previous studies, the process before being controlled was assumed to be a spectrally one-sided L\'evy process, however in this paper we use a L\'evy process that may have both positive and negative jumps. In the main theorem, we show that a refraction--reflection strategy is an optimal strategy. We also mention the existence and uniqueness of solutions of the stochastic differential equations that define refracted L\'evy processes.

[9]  arXiv:2110.09638 (cross-list from cs.GT) [pdf, other]
Title: Repeated Games, Optimal Channel Capture, and Open Problems for Slotted Multiple Access
Authors: Michael J. Neely
Comments: 19 pages
Subjects: Computer Science and Game Theory (cs.GT); Optimization and Control (math.OC)

This paper revisits a classical problem of slotted multiple access with success, idle, and collision events on each slot. First, results of a 2-user multiple access game are reported. The game was conducted at the University of Southern California over multiple semesters and involved competitions between student-designed algorithms. An algorithm called 4-State was a consistent winner. This algorithm is analyzed and shown to have an optimal expected score when competing against an independent version of itself. The structure of 4-State motivates exploration of the open question of how to minimize the expected time to capture the channel for a $n$-user situation. It is assumed that the system delivers perfect feedback on the number of users who transmitted at the end of each slot. An efficient algorithm is developed and conjectured to have an optimal expected capture time for all positive integers $n$. Optimality is proven in the special cases $n \in \{1, 2, 3, 4, 6\}$ using a novel analytical technique that introduces virtual users with enhanced capabilities.

[10]  arXiv:2110.09908 (cross-list from math.PR) [pdf, other]
Title: On random walks and switched random walks on homogeneous spaces
Comments: 23 pages, 3 figures
Subjects: Probability (math.PR); Optimization and Control (math.OC)

We prove new mixing rate estimates for the random walks on homogeneous spaces determined by a probability distribution on a finite group $G$. We introduce the switched random walk determined by a finite set of probability distributions on $G$, prove that its long-term behavior is determined by the Fourier joint spectral radius of the distributions and give hermitian sum-of-squares algorithms for the effective estimation of this quantity.

[11]  arXiv:2110.09993 (cross-list from cs.DC) [pdf, other]
Title: A Unified and Refined Convergence Analysis for Non-Convex Decentralized Learning
Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Optimization and Control (math.OC)

We study the consensus decentralized optimization problem where the objective function is the average of $n$ agents private non-convex cost functions; moreover, the agents can only communicate to their neighbors on a given network topology. The stochastic online setting is considered in this paper where each agent can only access a noisy estimate of its gradient. Many decentralized methods can solve such problems including EXTRA, Exact-Diffusion/D$^2$, and gradient-tracking. Unlike the famed $\small \text{DSGD}$ algorithm, these methods have been shown to be robust to the heterogeneity of the local cost functions. However, the established convergence rates for these methods indicate that their sensitivity to the network topology is worse than $\small \text{DSGD}$. Such theoretical results imply that these methods can perform much worse than $\small \text{DSGD}$ over sparse networks, which, however, contradicts empirical experiments where $\small \text{DSGD}$ is observed to be more sensitive to the network topology.
In this work, we study a general stochastic unified decentralized algorithm ($\small\textbf{SUDA}$) that includes the above methods as special cases. We establish the convergence of $\small\textbf{SUDA}$ under both non-convex and the Polyak-Lojasiewicz condition settings. Our results provide improved network topology dependent bounds for these methods (such as Exact-Diffusion/D$^2$ and gradient-tracking) compared with existing literature. Moreover, our result shows that these method are less sensitive to the network topology compared to $\small \text{DSGD}$, which agrees with numerical experiments.

[12]  arXiv:2110.10079 (cross-list from math.CA) [pdf, ps, other]
Title: Sparse non-SOS Putinar-type Positivstellensätze
Subjects: Classical Analysis and ODEs (math.CA); Algebraic Geometry (math.AG); Optimization and Control (math.OC)

Recently, non-SOS Positivstellens\"atze for polynomials on compact semialgebraic sets, following the general form of Schm\"{u}dgen's Positivstellensatz, have been derived by appropriately replacing the SOS polynomials with other classes of polynomials. An open question in the literature is how to obtain similar results following the general form of Putinar's Positivstellensatz. Extrapolating the algebraic geometry tools used to obtain this type of result in the SOS case fails to answer this question, because algebraic geometry methods strongly use hallmark properties of the class of SOS polynomials, such as closure under multiplication and closure under composition with other polynomials. In this article, using a new approach, we show the existence of Putinar-type Positivstellens\"atze that are constructed using non-SOS classes of non-negative polynomials, such as SONC, SDSOS and DSOS polynomials. Even not necessarily non-negative classes of polynomials such as sums of arithmetic-mean/geometric-mean polynomials could be used. Furthermore, we show that these certificates can be written with inherent sparsity characteristics. Such characteristics can be further exploited when the sparsity structure of both the polynomial whose non-negativity is being certified and the polynomials defining the semialgebraic set of interest are known. In contrast with related literature focused on exploiting sparsity in SOS Positivstellens\"atze, these latter results show how to exploit sparsity in a more general setting in which non-SOS polynomials are used to construct the Positivstellens\"atze.

[13]  arXiv:2110.10093 (cross-list from eess.IV) [pdf, other]
Title: Stochastic Primal-Dual Deep Unrolling Networks for Imaging Inverse Problems
Authors: Junqi Tang
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Optimization and Control (math.OC)

In this work we present a new type of efficient deep-unrolling networks for solving imaging inverse problems. Classical deep-unrolling methods require full forward operator and its adjoint across each layer, and hence can be computationally more expensive than other end-to-end methods such as FBP-ConvNet, especially in 3D image reconstruction tasks. We propose a stochastic (ordered-subsets) extension of the Learned Primal-Dual (LPD) which is the state-of-the-art unrolling network. In our unrolling network, we only use a subset of the forward and adjoint operator, to achieve computational efficiency. We consider 3 ways of training the proposed network to cope with different scenarios of the availability of the training data, including (1) supervised training on paired data, (2) unsupervised adversarial training which enable us to train the network without paired ground-truth data, (3) equivariant self-supervised training approach, which utilizes equivariant structure which is prevalent in many imaging applications, and only requires measurement data. Our numerical results demonstrate the effectiveness of our approach in X-ray CT imaging task, showing that our networks achieve similar reconstruction accuracies as the full-batch LPD, while require only a fraction of the computation.

[14]  arXiv:2110.10116 (cross-list from cs.LG) [pdf, ps, other]
Title: On the Global Convergence of Momentum-based Policy Gradient
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC)

Policy gradient (PG) methods are popular and efficient for large-scale reinforcement learning due to their relative stability and incremental nature. In recent years, the empirical success of PG methods has led to the development of a theoretical foundation for these methods. In this work, we generalize this line of research by studying the global convergence of stochastic PG methods with momentum terms, which have been demonstrated to be efficient recipes for improving PG methods. We study both the soft-max and the Fisher-non-degenerate policy parametrizations, and show that adding a momentum improves the global optimality sample complexity of vanilla PG methods by $\tilde{\mathcal{O}}(\epsilon^{-1.5})$ and $\tilde{\mathcal{O}}(\epsilon^{-1})$, respectively, where $\epsilon>0$ is the target tolerance. Our work is the first one that obtains global convergence results for the momentum-based PG methods. For the generic Fisher-non-degenerate policy parametrizations, our result is the first single-loop and finite-batch PG algorithm achieving $\tilde{O}(\epsilon^{-3})$ global optimality sample complexity. Finally, as a by-product, our methods also provide general framework for analyzing the global convergence rates of stochastic PG methods, which can be easily applied and extended to different PG estimators.

[15]  arXiv:2110.10133 (cross-list from cs.LG) [pdf, other]
Title: Locally Differentially Private Reinforcement Learning for Linear Mixture Markov Decision Processes
Comments: 25 pages, 2 figures
Subjects: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Optimization and Control (math.OC); Machine Learning (stat.ML)

Reinforcement learning (RL) algorithms can be used to provide personalized services, which rely on users' private and sensitive data. To protect the users' privacy, privacy-preserving RL algorithms are in demand. In this paper, we study RL with linear function approximation and local differential privacy (LDP) guarantees. We propose a novel $(\varepsilon, \delta)$-LDP algorithm for learning a class of Markov decision processes (MDPs) dubbed linear mixture MDPs, and obtains an $\tilde{\mathcal{O}}( d^{5/4}H^{7/4}T^{3/4}\left(\log(1/\delta)\right)^{1/4}\sqrt{1/\varepsilon})$ regret, where $d$ is the dimension of feature mapping, $H$ is the length of the planning horizon, and $T$ is the number of interactions with the environment. We also prove a lower bound $\Omega(dH\sqrt{T}/\left(e^{\varepsilon}(e^{\varepsilon}-1)\right))$ for learning linear mixture MDPs under $\varepsilon$-LDP constraint. Experiments on synthetic datasets verify the effectiveness of our algorithm. To the best of our knowledge, this is the first provable privacy-preserving RL algorithm with linear function approximation.

Replacements for Wed, 20 Oct 21

[16]  arXiv:1810.02429 (replaced) [pdf, other]
Title: Restarting Frank-Wolfe: Faster Rates Under Hölderian Error Bounds
Comments: Journal version
Subjects: Optimization and Control (math.OC)
[17]  arXiv:2012.05314 (replaced) [pdf, ps, other]
Title: Tropical complementarity problems and Nash equilibria
Subjects: Optimization and Control (math.OC)
[18]  arXiv:2101.00838 (replaced) [pdf, ps, other]
Title: Distributionally robust second-order stochastic dominance constrained optimization with Wasserstein ball
Subjects: Optimization and Control (math.OC)
[19]  arXiv:2105.12222 (replaced) [pdf, other]
Title: Projecting onto rectangular matrices with prescribed row and column sums
Subjects: Optimization and Control (math.OC)
[20]  arXiv:2110.06514 (replaced) [pdf, ps, other]
Title: Convexity of sets and functions on the hyperbolic space
Subjects: Optimization and Control (math.OC)
[21]  arXiv:2110.07269 (replaced) [pdf, other]
Title: Momentum-Based Nash Set Seeking over Networks via Multi-Time Scale Hybrid Dynamic Inclusions
Comments: 19 pages, 9 figures
Subjects: Optimization and Control (math.OC)
[22]  arXiv:2110.08471 (replaced) [pdf, other]
Title: Fast Projection onto the Capped Simplex withApplications to Sparse Regression in Bioinformatics
Comments: 12 pages, 5 figures
Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG); Genomics (q-bio.GN)
[23]  arXiv:2001.02958 (replaced) [pdf, other]
Title: Shape optimization of a weighted two-phase Dirichlet eigenvalue
Authors: Idriss Mazari (CEREMADE), Grégoire Nadin (LJLL), Yannick Privat (TONUS, IUF, IRMA)
Subjects: Analysis of PDEs (math.AP); Optimization and Control (math.OC)
[24]  arXiv:2102.13566 (replaced) [pdf, other]
Title: Sparse approximation in learning via neural ODEs
Comments: 24 pages, 5 figures
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
[25]  arXiv:2106.02899 (replaced) [pdf, ps, other]
Title: $L^\infty$-estimates in optimal transport for non quadratic costs
Comments: A typo is corrected
Subjects: Analysis of PDEs (math.AP); Optimization and Control (math.OC)
[26]  arXiv:2108.12547 (replaced) [pdf, other]
Title: Self-fulfilling Bandits: Dynamic Selection in Algorithmic Decision-making
Comments: Main Body: 30 pages, 6 figures; Supplemental Material: 26 pages
Subjects: Econometrics (econ.EM); Machine Learning (cs.LG); Optimization and Control (math.OC); Methodology (stat.ME); Machine Learning (stat.ML)
[27]  arXiv:2109.12836 (replaced) [pdf, ps, other]
Title: Optimal control of a first order Fokker-Planck equation with reaction term and density constraints
Authors: Adrien Seguret (EDF R&D OSIRIS, CEREMADE)
Subjects: Analysis of PDEs (math.AP); Optimization and Control (math.OC)
[28]  arXiv:2110.08440 (replaced) [pdf, other]
Title: Online Target Q-learning with Reverse Experience Replay: Efficiently finding the Optimal Policy for Linear MDPs
Comments: Under Review, V2 has updated acknowledgements
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC)
[ total of 28 entries: 1-28 ]
[ showing up to 2000 entries per page: fewer | more ]

Disable MathJax (What is MathJax?)

Links to: arXiv, form interface, find, math, recent, 2110, contact, help  (Access key information)