Article
Chemistry, Analytical
Mariusz Frackiewicz, Henryk Palus
Summary: This paper proposes three methods for color quantization of superpixel images, showing significantly reduced computation time and high quality quantized images, although the image quality is slightly lower compared to pixel methods.
Article
Physics, Multidisciplinary
John C. Napp, Rolando L. La Placa, Alexander M. Dalzell, Fernando G. S. L. Brandao, Aram W. Harrow
Summary: The advantage of quantum computation over classical computation lies in its ability to efficiently simulate situations that are difficult to simulate on classical computers. We have proven that certain families of random circuits can be approximately simulated on classical computers in linear time, even though they have the capability of universal quantum computation and are hard to simulate exactly. We propose two classical simulation algorithms and provide numerical and analytical evidence for their efficiency.
Article
Astronomy & Astrophysics
Andreas Nygaard, Emil Brinch Holm, Steen Hannestad, Thomas Tram
Summary: The paper discusses the significance of profile likelihoods in cosmology and proposes a method to accommodate their computational requirements using neural networks and an optimization algorithm. The approach allows for high-precision parameter estimation and the production of triangle plots commonly associated with Bayesian marginalization.
JOURNAL OF COSMOLOGY AND ASTROPARTICLE PHYSICS
(2023)
Article
Multidisciplinary Sciences
Tom Edinburgh, Ari Ercole, Stephen Eglen
Summary: Multilevel linear models are used to flexibly model complex data with different levels of stratification. In Bayesian analysis, model comparison is typically done using model evidence or Bayes factor, but computation of these quantities is often impossible. In this paper, a method is proposed to estimate the log model evidence by marginalizing over non-variance parameters, which reduces the dimensionality of Monte Carlo sampling and leads to more consistent estimates.
Article
Statistics & Probability
Samuel Duffield, Sumeetpal S. Singh
Summary: This letter generalizes Ensemble Kalman inversion techniques to general Bayesian models, where previously they were limited to additive Gaussian likelihoods. The setting is challenging as the likelihood can be sampled from, but its density may not be evaluated.
STATISTICS & PROBABILITY LETTERS
(2022)
Article
Computer Science, Information Systems
Keiichi Iwamura, Ahmad Akmal Aminuddin Mohd Kamal
Summary: In this study, a new secure computation method based on a client-server model is proposed, which uses (k, n) threshold secret sharing for computation without communication, even during multiplication. Unlike conventional methods, our approach concentrates communication in the preprocessing phase, resulting in faster overall processing. Extensive security analysis and experimental simulation are conducted to validate the proposed method.
Article
Chemistry, Physical
Benchen Huang, Nan Sheng, Marco Govoni, Giulia Galli
Summary: We present a computational protocol for quantum simulations of fermionic Hamiltonians, using a qubit-efficient encoding scheme and a modified qubit-coupled cluster ansatz. The protocol improves the scaling of circuit gate counts and the required number of qubits, as well as increasing resilience to noise. We demonstrate the effectiveness of the protocol by simulating spin defects in diamond and 4H silicon carbide.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Biochemical Research Methods
Almut Heinken, Ines Thiele
Summary: This study improved the computation speed and scalability of the constraint-based microbiome modeling and analysis approach, and added additional functionalities for analysis and visualization.
Article
Computer Science, Information Systems
Mahdi Zahedi, Taha Shahroodi, Stephan Wong, Said Hamdioui
Summary: This paper proposes an energy and area-efficient method for arithmetic signed matrix multiplications based on memristor-based computation-in-memory (CIM) engines. The method includes mapping the signed operands without sign extension on the 1T1R crossbar and augmenting the periphery with customized circuits for shift and accumulate operations. Simulations for different case studies show that the proposed method achieves up to 8x energy-saving and 3x area-saving compared to other CIM-based approaches.
Article
Physics, Particles & Fields
N. Berger
Summary: This paper presents a simplified likelihood framework for LHC experimental results, which allows for easy reuse, reinterpretation, and combination of the data. The framework is based on the widely used HistFactory format but simplifies the consideration of systematic uncertainties to linear order only. This simplification greatly improves computing performance while accurately capturing non-Gaussian effects and correlated uncertainties.
JOURNAL OF HIGH ENERGY PHYSICS
(2023)
Article
Computer Science, Hardware & Architecture
Azam Ghanbari, Mehdi Modarressi
Summary: This paper introduces CORN-C, a computation reuse-aware accelerator for convolutional neural networks. By eliminating repetitive computations, CORN-C achieves energy and latency reduction.
JOURNAL OF SYSTEMS ARCHITECTURE
(2022)
Article
Multidisciplinary Sciences
A. H. Homid, M. Abdel-Aty, M. Qasymeh, H. Eleuch
Summary: This work demonstrates that using trapped ultracold atoms as a platform for quantum gate circuits and algorithms can offer better performance compared to previously reported approaches.
SCIENTIFIC REPORTS
(2021)
Article
Physics, Fluids & Plasmas
Sheng-Chen Liu, Lin Cheng, Gui-Zhong Yao, Ying-Xiang Wang, Liang-You Peng
Summary: Almost every quantum circuit in current stage relies on two-qubit gates, which are crucial for quantum computing in any platform. In trapped-ion systems, entangling gates based on Molmer-Sorensen schemes are widely used, utilizing the collective motional modes of ions and laser-controlled internal states as qubits. The key to achieving high-fidelity and robust gates is minimizing entanglement between qubits and motional modes under various sources of errors. In this work, an efficient numerical method is proposed to search for high-quality solutions for phase-modulated pulses, which solves the problem through a combination of linear algebra and quadratic equations. The method demonstrates effectiveness up to 60 ions and overcomes convergence issues, meeting the gate design needs in current trapped-ion experiments.
Article
Computer Science, Information Systems
Tao Wang, Zhusen Liu, Zhaoyang Han, Lu Zhou
Summary: In the era of big data, securely combining private data owned by different companies or organizations is crucial for making correct decisions. We have designed a secure and efficient decision-making scheme that allows clients to outsource data and computations to cloud servers while ensuring integrity, confidentiality, and correctness.
Article
Chemistry, Multidisciplinary
Yu Fu, Li Li, Yujin Hu
Summary: This study proposes an efficient modal modification method for topology optimization of large-scale multi-physics models. Through the construction of Krylov reduced-basis vectors and a master-slave pattern parallel method, the computational cost is reduced and the efficiency is improved.
APPLIED SCIENCES-BASEL
(2022)
Article
Operations Research & Management Science
Charles Audet, Kwassi Joseph Dzahini, Michael Kokkolaras, Sebastien Le Digabel
Summary: The proposed StoMADS algorithm is a stochastic extension of the MADS algorithm designed for deterministic blackbox optimization, aiming to optimize the objective function f under the influence of random noise. By utilizing random estimates of function values obtained from stochastic observations, the algorithm aims to converge to a Clarke stationary point of f.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2021)
Article
Operations Research & Management Science
Jean Bigeon, Sebastien Le Digabel, Ludovic Salomon
Summary: The research introduces a new method for multiobjective derivative-free optimization, maintaining a list of non-dominated points converging to a (local) Pareto set. Computational experiments suggest competitive performance compared to state-of-the-art algorithms.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2021)
Article
Operations Research & Management Science
Stephane Alarie, Charles Audet, Aimen E. Gheribi, Michael Kokkolaras, Sebastien Le Digabel
Summary: This article reviews the applications of direct search optimization methods in blackbox optimization over the past twenty years, with a focus on the Mesh Adaptive Direct Search (MAnS) derivative-free optimization algorithm and its applications in energy, materials science, and computational engineering design. The versatility of MAnS and the evolution of its accompanying software NOMAD as a standard tool for blackbox optimization are highlighted through a range of applications in various fields of science and engineering.
EURO JOURNAL ON COMPUTATIONAL OPTIMIZATION
(2021)
Article
Operations Research & Management Science
Charles Audet, Frederic Messine, Jordan Ninin
Summary: This paper studies the problem of numerically certifying the proximity between a conjectured set and the Pareto set, using a computer assisted proof methodology. The conjectured set can be directly given or parameterized, and the verification question is formulated as a global optimization problem. The effectiveness of the method is illustrated through a class of extremal problems over convex polygons in the plane.
JOURNAL OF GLOBAL OPTIMIZATION
(2022)
Article
Mathematics
Christian Bingane, Charles Audet
Summary: In this study, a family of convex equilateral small n-gons was constructed and their perimeters were shown to be close to the maximal perimeter. Furthermore, it was proven that the previously known equilateral hexadecagon is not optimal for the case of n = 16.
ARCHIV DER MATHEMATIK
(2022)
Correction
Operations Research & Management Science
Charles Audet, Frederic Messine, Jordan Ninin
JOURNAL OF GLOBAL OPTIMIZATION
(2022)
Article
Computer Science, Software Engineering
Kwassi Joseph Dzahini, Michael Kokkolaras, Sebastien Le Digabel
Summary: The StoMADS-PB algorithm is proposed for constrained stochastic blackbox optimization, which deals with objective and constraint functions provided by a noisy blackbox. The method utilizes estimates and probabilistic bounds for constraint violations, allowing infeasible solutions. With the use of Clarke nonsmooth calculus and martingale theory, convergence results for the objective and violation function are proven with probability one.
MATHEMATICAL PROGRAMMING
(2023)
Article
Mathematics, Applied
Christian Bingane, Charles Audet
Summary: This paper solves the problem of the maximal width of equilateral small polygons and finds the optimal equilateral small octagon. It also proposes a family of equilateral small n-gons with widths approximating the maximal width.
MATHEMATICS OF COMPUTATION
(2022)
Article
Computer Science, Software Engineering
Charles Audet, Sebastien Le Digabel, Viviane Rochon Montplaisir, Christophe Tribes
Summary: NOMAD is a state-of-the-art software package for optimizing blackbox problems, with the latest version being NOMAD 4 which features a complete redesign with a new architecture providing more flexible code and added functionalities. NOMAD 4 introduces algorithmic components to simplify the implementation of complex algorithms, and utilizes parallelism to maximize optimization performance.
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE
(2022)
Article
Operations Research & Management Science
Charles Audet, Sebastien Le Digabel, Renaud Saltet
Summary: This research introduces a new approach to solving blackbox optimization problems by building ensembles of surrogates and quantifying the uncertainty of their predictions, which allows for optimization at a lower computational cost. Computational experiments demonstrate that this method achieves higher precision and lower computational effort compared to traditional stochastic models in solving expensive simulation problems.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2022)
Article
Mathematics, Applied
Charles Audet, Alain Batailly, Solene Kojtych
Summary: The design of key nonlinear systems often requires the use of expensive blackbox simulations presenting inherent discontinuities. This paper proposes the DiscoMADS algorithm to solve the optimization problem by building inner approximations and successfully avoids unsafe regions in the design space.
SIAM JOURNAL ON OPTIMIZATION
(2022)
Article
Operations Research & Management Science
Stephane Alarie, Charles Audet, Paulin Jacquot, Sebastien Le Digabel
Summary: This study proposes two strategies to optimize the evaluation process of constraint functions in blackbox optimization and validates them through experiments.
OPERATIONS RESEARCH LETTERS
(2022)