Article
Computer Science, Theory & Methods
Sandra Catalan, Francisco D. Igual, Jose R. Herrero, Rafael Rodriguez-Sanchez, Enrique S. Quintana-Orti
Summary: This paper proposes a methodology to address programmability issues in new-generation shared-memory NUMA architectures. By using dense matrix factorizations and matrix inversion as a use case, the methodology achieves performance portability across different NUMA configurations through multi-domain implementations and hybrid task- and loop-level parallelization. The experimentation validates the proposal on two modern architectures and demonstrates competitive performance with state-of-the-art message-passing implementations.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2023)
Article
Automation & Control Systems
Agniva Chowdhury, Gregory Dexter, Palma London, Haim Avron, Petros Drineas
Summary: Linear programming is a useful tool that has been successfully applied in various fields. This paper focuses on the special case where the number of variables is much larger than the number of constraints and proposes a preconditioning technique to ensure the convergence and optimization of the algorithm.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Victor Wen-zhe Yu, Jonathan Moussa, Pavel Kus, Andreas Marek, Peter Messmer, Mina Yoon, Hermann Lederer, Volker Blum
Summary: This work presents GPU-oriented optimizations for the ELPA2 eigensolver, demonstrating superior performance through benchmarking on two hybrid CPU-GPU architectures.
COMPUTER PHYSICS COMMUNICATIONS
(2021)
Article
Computer Science, Information Systems
Jeongeun Kim, Won Sik Jeong, Youngwoo Jeong, Seung Eun Lee
Summary: This paper proposes a stochastic computing architecture to address latency issues by using parallel linear feedback shift registers (LFSRs). The proposed architecture reduces the latency in the stochastic sequence generation process without sacrificing accuracy. Additionally, it achieves area efficiency by reducing flip-flops and lookup tables compared to architectures using shared LFSRs or multiple LFSRs.
Article
Engineering, Electrical & Electronic
Felipe D. R. Machado, Andre Luiz Diniz, Carmen L. T. Borges, Lilian C. Brandao
Summary: Optimizing power generation planning is crucial for cost-effectiveness and meeting power demand, and the use of parallel computing methods can significantly enhance the efficiency of solving this problem.
ELECTRIC POWER SYSTEMS RESEARCH
(2021)
Article
Engineering, Multidisciplinary
Franco Dassi, Stefano Zampini, S. Scacchi
Summary: This paper introduces the Virtual Element Method (VEM) and its linear solver for solving three-dimensional elliptic equations. The proposed Balancing Domain Decomposition by Constraints (BDDC) preconditioner effectively controls the condition number of the system. Experimental results confirm the reliability and adaptability of the method.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Multidisciplinary Sciences
Kin Keung Lai, Mohd Hassan, Jitendra Kumar Maurya, Sanjeev Kumar Singh, Shashi Kant Mishra
Summary: This paper considers convex multiobjective optimization problems with equality and inequality constraints in real Banach space, establishing saddle point necessary and sufficient Pareto optimality conditions under certain constraint qualifications, and also discussing second order symmetric duality for nonlinear multiobjective mixed integer programs for arbitrary cones. The study also relates to necessary and sufficient optimality conditions for vector equilibrium problems on Hadamard manifolds by Ruiz-Garzon et al. in 2019.
Article
Computer Science, Theory & Methods
Gabriele Mencagli, Massimo Torquati, Andrea Cardaci, Alessandra Fais, Luca Rinaldi, Marco Danelutto
Summary: Nowadays, there is a growing focus on Stream Processing Systems (SPSs) for scale-up machines, where some systems still use the continuous model for low latency while others optimize throughput with batching approaches. The approach presented in the text aims to design a runtime system of SPSs targeting multicores to optimize throughput and latency, using building blocks for easy and compositional optimizations.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2021)
Article
Computer Science, Interdisciplinary Applications
Ragheb Rahmaniani, Teodor Gabriel Crainic, Michel Gendreau, Walter Rei
Summary: Benders decomposition (BD) is a popular solution algorithm for stochastic integer programs. However, existing parallelization methods often suffer from inefficiencies. This paper proposes an asynchronous parallel BD method and demonstrates its effectiveness through numerical studies and performance enhancement strategies.
COMPUTERS & OPERATIONS RESEARCH
(2024)
Article
Automation & Control Systems
Zhiyong Yu, Baokai Zhang, Feng Zhang
Summary: This paper investigates a type of two-person zero-sum linear-quadratic stochastic differential game with jumps and establishes a closed-loop formulation in mixed-strategy-law form for the saddle points of the game.
INTERNATIONAL JOURNAL OF CONTROL
(2022)
Article
Operations Research & Management Science
Thanakorn Khamvilai, Louis Sutter, Philippe Baufreton, Francois Neumann, Eric Feron
Summary: This research introduces an online decentralized allocation algorithm for safety-critical applications on parallel computing architectures, utilizing an abstract graph representation of the architecture and computational units, solving the allocation problem through Integer Linear Programming, achieving decentralization through redundancy, improving system reliability, and demonstrating experimental application in avionics systems.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2021)
Article
Management
Djellouli Younes, Hamadou Sarah, Chaabane Djamal
Summary: This paper introduces a modified sequential version method for optimizing a linear function over an integer efficient set, as well as a new exact parallel algorithm. The performance of parallel programming in this context is demonstrated through various instances with different sizes, where a finite monotonous sequence of values for the main criterion is built in a reasonable amount of CPU execution time.
INTERNATIONAL TRANSACTIONS IN OPERATIONAL RESEARCH
(2022)
Article
Computer Science, Software Engineering
Bernard Knueven, David Mildebrath, Christopher Muir, John D. D. Siirola, Jean-Paul Watson, David L. L. Woodruff
Summary: mpi-sppy is an open source package that provides efficient and scalable parallelization for solving large stochastic programming problems in minutes. It introduces a novel software architecture for accelerating convergence, and is written in Python with the widely used Pyomo library.
MATHEMATICAL PROGRAMMING COMPUTATION
(2023)
Article
Mathematics
Alberto Manzano, Daniele Musso, Alvaro Leitao, Andres Gomez, Carlos Vazquez, Gustavo Ordonez, Maria R. Nogueiras
Summary: The framework for designing quantum algorithms relies on quantum matrix and three quasi-independent modules. Loading and read-out modules are briefly discussed, while more in-depth analysis is done on the arithmetic module. Examples regarding manipulation of generic oracles are provided, hinting at potential applications.
Article
Computer Science, Interdisciplinary Applications
Ambros Gleixner, Leona Gottwald, Alexander Hoen
Summary: This paper presents PAPILO, a new C++ header-only library that provides a large set of pre solving routines for MIP and linear programming problems. PAPILO aims to optimize computational performance by parallelization and support for multiprecision arithmetic.
INFORMS JOURNAL ON COMPUTING
(2023)
Article
Engineering, Electrical & Electronic
Johannes J. Brust, Mihai Anitescu
Summary: This article analyzes the convergence conditions for a fixed point iteration method used in optimal power flow problems with chance constraints, and presents numerical experiments including for large IEEE networks.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Daniel Adrian Maldonado, Emil M. Constantinescu, Hong Zhang, Vishwas Rao, Mihai Anitescu
Summary: This work introduces a novel technique to compute extreme trajectories of power system dynamic simulations using trust-region optimization algorithm and second-order trajectory sensitivities. The method overcomes limitations of previous sensitivity-based techniques in approximating bounds of trajectories in nonlinear scenarios, showing accuracy and scalability in numerical experiments.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Anirudh Subramanyam, Jacob Roth, Albert Lam, Mihai Anitescu
Summary: This paper proposes a reliability-aware approach for power dispatch, which aims to reduce the probability of cascades by considering the failure potential. By utilizing a probability model, the proposed method can effectively decrease the risk of blackouts in power transmission systems.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2022)
Article
Operations Research & Management Science
Miles Lubin, Juan Pablo Vielma, Ilias Zadik
Summary: This study examines which sets can be represented exactly as feasible regions of mixed-integer convex optimization problems, establishing a complete characterization for the mixed-binary case and necessary conditions for the general case. The results provide the first nonrepresentability findings for various nonconvex sets and suggest a clear separation between the modeling power of mixed-integer linear and mixed-integer convex optimization, particularly in the case of subsets of natural numbers. The study also highlights the necessity of unbounded integer variables for modeling unbounded sets in the case of compact sets.
MATHEMATICS OF OPERATIONS RESEARCH
(2022)
Article
Engineering, Electrical & Electronic
Bendong Tan, Junbo Zhao, Nan Duan, Daniel Adrian Maldonado, Yingchen Zhang, Hongming Zhang, Mihai Anitescu
Summary: This paper proposes a distributed frequency divider to estimate power system bus frequency with a limited number of PMUs while considering the inertial contributions from DFIGs. The proposed method reformulates the original frequency divider by modeling the contributions of DFIGs, allowing estimation of local bus frequencies in a distributed manner. The numerical results on real power systems demonstrate the effectiveness and robustness of the proposed method.
IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS
(2022)
Article
Physics, Multidisciplinary
Puspa Upreti, Matthew Krogstad, Charlotte Haley, Mihai Anitescu, Vishwas Rao, Lekh Poudel, Omar Chmaissem, Stephan Rosenkranz, Raymond Osborn
Summary: The paper investigates structural correlations in the quasiskutterudites using single crystal x-ray diffraction and finds temperature-independent local atomic displacements below the transition, which persist to well above the transition, indicating order-disorder behavior.
PHYSICAL REVIEW LETTERS
(2022)
Article
Mathematics, Applied
Sungho Shin, Mihai Anitescu, Victor M. Zavala
Summary: This study investigates the solution sensitivity of nonlinear programs based on graph structures and presents the exponential decay result of solution sensitivity with respect to distance. This result holds under the strong second-order sufficiency condition and the linear independence constraint qualification, and is illustrated with numerical examples.
SIAM JOURNAL ON OPTIMIZATION
(2022)
Article
Computer Science, Software Engineering
Sen Na, Mihai Anitescu, Mladen Kolar
Summary: In this paper, we propose a stochastic algorithm based on sequential quadratic programming (SQP) for solving nonlinear optimization problems with a stochastic objective and deterministic equality constraints. We introduce a differentiable exact augmented Lagrangian as the merit function and incorporate a stochastic line search procedure to adaptively select the random stepsizes. The global almost sure convergence for both non-adaptive and adaptive SQP methods is established. Numerical experiments demonstrate the superiority of the adaptive algorithm.
MATHEMATICAL PROGRAMMING
(2023)
Article
Computer Science, Software Engineering
David Applegate, Oliver Hinder, Haihao Lu, Miles Lubin
Summary: This paper utilizes the sharpness of primal-dual formulations in linear programming to achieve linear convergence with restarts, improving the accuracy of solution.
MATHEMATICAL PROGRAMMING
(2023)
Article
Automation & Control Systems
Sen Na, Mihai Anitescu
Summary: We have developed an online, lag-L, model predictive control (MPC) algorithm that solves discrete-time, equality-constrained, nonlinear dynamic programs with one Newton step per horizon. Based on recent sensitivity analysis results, we have proven that this approach exhibits a behavior called superconvergence, where the tracking error with respect to the full-horizon solution is stable for successive horizon shifts and decreases with increasing shift order to a minimum value that decays exponentially in the length of the receding horizon. We have shown that the algorithmic error achieved quadratic convergence through Newton's method, while the perturbation error decayed exponentially with the lag between consecutive receding horizons. Overall, this approach induces local exponential convergence in terms of the receding horizon length for suitable values of L. Numerical experiments have validated our theoretical findings.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2023)
Article
Operations Research & Management Science
Francois Pacaud, Sungho Shin, Michel Schanen, Daniel Adrian Maldonado, Mihai Anitescu
Summary: The interior-point method (IPM) is widely used in nonlinear programming, and its performance depends on the linear solver employed to solve the Karush-Kuhn-Tucker (KKT) system. We propose a novel reduced-space IPM algorithm that condenses the KKT system into a dense matrix proportional to the number of degrees of freedom. Two variants of the reduced-space method are derived, and computations are accelerated on GPUs. Extensive numerical results show that the reduced-space algorithms outperform Knitro and a hybrid full-space IPM algorithm when the relative number of degrees of freedom decreases.
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS
(2023)
Article
Computer Science, Software Engineering
Ilias Zadik, Miles Lubin, Juan Pablo Vielma
Summary: We investigate the structural geometric properties of mixed-integer convex representable (MICP-R) sets and compare them with the class of mixed-integer linear representable (MILP-R) sets. We provide examples of MICP-R sets that are countably infinite unions of convex sets with countably infinitely many different recession cones, and countably infinite unions of polytopes with different shapes. These examples highlight the differences between MICP-R sets and MILP-R sets.
MATHEMATICAL PROGRAMMING
(2023)
Article
Computer Science, Software Engineering
Sen Na, Mihai Anitescu, Mladen Kolar
Summary: We propose an active-set stochastic sequential quadratic programming (StoSQP) algorithm for solving nonlinear optimization problems with a stochastic objective and deterministic constraints. The algorithm uses a differentiable exact augmented Lagrangian as the merit function and adaptively selects penalty parameters. It has been shown to have global convergence and outperforms previous work in terms of nonlinear inequality constraints and sample complexity.
MATHEMATICAL PROGRAMMING
(2023)
Article
Computer Science, Software Engineering
Miles Lubin, Oscar Dowson, Joaquim Dias Garcia, Joey Huchette, Benoit Legat, Juan Pablo Vielma
Summary: JuMP is an algebraic modeling language in Julia, which can be used to model various types of optimization problems and handles the low-level details of communicating with solvers. After almost 10 years of development, JuMP 1.0 was released in March 2022.
MATHEMATICAL PROGRAMMING COMPUTATION
(2023)
Article
Energy & Fuels
Amanda Lenzi, Julie Bessac, Mihai Anitescu
Summary: This study proposes statistical models for anticipating disturbances in the frequency response at multiple locations simultaneously and issues real-time event recommendations based on predefined loss functions. The case study shows that the proposed spatiotemporal model better captures uncertainty and increases the probability of identifying large frequency excursions, reducing risks for the system operator.
IEEE OPEN ACCESS JOURNAL OF POWER AND ENERGY
(2022)