Article
Mathematics
Johny Pambabay-Calero, Sergio Bauz-Olvera, Ana Nieto-Librero, Ana Sanchez-Garcia, Puri Galindo-Villardon
Summary: Models implemented in statistical software for precision analysis of diagnostic tests include random-effects modeling and hierarchical regression, but calculating the covariance between sensitivity and specificity is challenging when the random effect is zero. Copulas are used as an alternative method, and posterior values are estimated using Markov chain Monte Carlo methods.
Article
Multidisciplinary Sciences
Jan Kessler, Francesco Calcavecchia, Thomas D. Kuehne
Summary: Inspired by the universal approximation theorem and the widespread adoption of artificial neural network techniques, feed-forward neural networks are proposed as a general purpose trial wave function for quantum Monte Carlo simulations of continuous many-body systems. The accuracy of the trial wave functions was demonstrated by studying an exactly solvable model system of two trapped interacting particles and the hydrogen dimer. The whole many-body wave function can be represented by a neural network for simple model systems, while the antisymmetry condition of non-trivial fermionic systems is incorporated by means of a Slater determinant.
ADVANCED THEORY AND SIMULATIONS
(2021)
Article
Mathematics, Applied
Andreas Frommer, Mostafa Nasr Khalil, Gustavo Ramirez-Hidalgo
Summary: This paper investigates a multilevel Monte Carlo approach for estimating the trace of a matrix function. By utilizing a multigrid hierarchy, the variance is significantly reduced, allowing for higher precision estimation with less effort.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2022)
Article
Computer Science, Interdisciplinary Applications
Bert Mortier, Pieterjan Robbe, Martine Baelmans, Giovanni Samaey
Summary: We have developed a novel multilevel asymptotic-preserving Monte Carlo method, called Multilevel Kinetic-Diffusion Monte Carlo (ML-KDMC), for simulating the kinetic Boltzmann transport equation. By incorporating this method within a Multilevel Monte Carlo (MLMC) framework and utilizing a hierarchy of larger time step sizes, the simulation cost is further reduced. The ML-KDMC method outperforms the single-level KDMC method by several orders of magnitude, demonstrating its efficiency.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)
Article
Astronomy & Astrophysics
Eric R. Anschuetz, Lena Funcke, Patrick T. Komiske, Serhii Kryhin, Jesse Thaler
Summary: This paper introduces a method to enhance the performance of annealing algorithms by using degeneracy engineering, illustrated through the example of l(0)-norm regularization for sparse linear regression. The results show that degeneracy engineering substantially improves the annealing performance, motivating its application to various regularized optimization problems.
Article
Mathematics, Applied
Riccardo Tosi, Ramon Amela, Rosa M. Badia, Riccardo Rossi
Summary: The necessity of dealing with uncertainties is increasing in various fields of science and engineering. The proposed asynchronous framework for Monte Carlo and Multilevel Monte Carlo methods aims to improve computational efficiency by introducing a new level of parallelism between batches. This approach maintains the reliability of state-of-the-art techniques while enhancing computational efficiency.
JOURNAL OF SCIENTIFIC COMPUTING
(2021)
Article
Chemistry, Physical
Matthew Jura, Marvin Bishop
Summary: This paper computes the scattering functions of nine generations of ideal tri-functional comb and dendrimer polymers using novel graph techniques. The properties of polymers with 9 to 3069 branches are explored. The g-ratios and scattering functions indicate that as the number of branches increases, comb polymers behave more like linear polymers with half the number of branches, while dendrimers become more like spherical objects.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Automation & Control Systems
Ajay Jasra, Jeremy Heng, Yaxian Xu, Adrian N. Bishop
Summary: In this study, we consider a class of finite-time horizon nonlinear stochastic optimal control problem and propose a new Monte Carlo method. The method addresses the issue of variance growth by casting optimal control estimation as a smoothing problem for a state-space model and applying smoothing algorithms based on particle Markov chain Monte Carlo. To further reduce the cost, a multilevel Monte Carlo method is developed.
INTERNATIONAL JOURNAL OF CONTROL
(2022)
Article
Green & Sustainable Science & Technology
Jian Wang, Xiang Gao, Zhili Sun
Summary: The paper introduces a method called multilevel Monte Carlo (MLMC) for time-variant reliability analysis, aiming to enhance computational efficiency while maintaining accuracy and robustness. By discretizing the time interval using a geometric sequence of different timesteps and estimating the cumulative probability of failure with corrections from all levels, the method optimizes the number of random samples at each level to minimize computational complexity. Independently computed corrections at each level allow achieving accuracy at a lower cost compared to crude Monte Carlo simulation, while maintaining robustness to nonlinearity and dimensions.
Article
Mathematics, Applied
Zhijian He, Zhenghang Xu, Xiaoqun Wang
Summary: This paper proposes a new variational Bayes algorithm based on multilevel Monte Carlo and randomized quasi-Monte Carlo sampling, for handling inference problems with intractable likelihood functions. Compared to existing algorithms, this algorithm has better performance and convergence rate.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2022)
Article
Computer Science, Interdisciplinary Applications
Yian Chen, Jeremy Hoskins, Yuehaw Khoo, Michael Lindsey
Summary: We propose a novel approach for computing committor functions, which describe transitions of a stochastic process between metastable states. The approach parametrizes the committor function and solves the variational formulation of the backward Kolmogorov equation with linear time and memory complexity in high-dimensional settings. Numerical results demonstrate the effectiveness of the proposed method for high-dimensional problems.
JOURNAL OF COMPUTATIONAL PHYSICS
(2023)
Article
Mathematics, Applied
Martin Hutzenthaler, Tuan Anh Nguyen
Summary: This article introduces the full history recursive multilevel Picard approximation method for semilinear parabolic partial differential equations, which can overcome the curse of dimensionality and covers cases with locally monotone coefficient functions.
APPLIED NUMERICAL MATHEMATICS
(2022)
Article
Engineering, Multidisciplinary
M. Croci, K. E. Willcox, S. J. Wright
Summary: This paper extends MLBLUE method to multi-output forward UQ problems and presents new semidefinite programming formulations for their optimal setup. These formulations yield the optimal number of samples required and the optimal selection of low-fidelity models to use. The new multi-output formulations can be solved reliably and efficiently.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Mathematics, Applied
Alexander D. Gilbert, Robert Scheichl
Summary: This paper presents an efficient multilevel quasi-Monte Carlo (MLQMC) algorithm for computing the expectation of the smallest eigenvalue of an elliptic eigenvalue problem with stochastic coefficients. The algorithm combines multiple strategies, including multilevel variance reduction, quasi-Monte Carlo methods, reuse of eigenvector, and a two-grid discretization scheme, to accelerate the approximation process.
IMA JOURNAL OF NUMERICAL ANALYSIS
(2023)
Article
Mathematics, Applied
Alexander D. Gilbert, Robert Scheichl
Summary: In this paper, a multilevel quasi-Monte Carlo (MLQMC) method is presented for approximating the expectation of the minimal eigenvalue of an elliptic eigenvalue problem. The algorithm is based on discretizations of the spatial domain and truncations of the stochastic parameter domain, and randomly shifted lattice rules are used to approximate the expectations. The main contribution of this paper is the rigorous analysis of the error of this algorithm.
IMA JOURNAL OF NUMERICAL ANALYSIS
(2023)