-
Convergence rates for random feature neural network approximation in molecular dynamics
Authors:
Xin Huang,
Petr Plechac,
Mattias Sandberg,
Anders Szepessy
Abstract:
Random feature neural network approximations of the potential in Hamiltonian systems yield approximations of molecular dynamics correlation observables that have the expected error $\mathcal{O}\big((K^{-1}+J^{-1/2})^{\frac{1}{2}}\big)$, for networks with $K$ nodes using $J$ data points, provided the Hessians of the potential and the observables are bounded. The loss function is based on the least…
▽ More
Random feature neural network approximations of the potential in Hamiltonian systems yield approximations of molecular dynamics correlation observables that have the expected error $\mathcal{O}\big((K^{-1}+J^{-1/2})^{\frac{1}{2}}\big)$, for networks with $K$ nodes using $J$ data points, provided the Hessians of the potential and the observables are bounded. The loss function is based on the least squares error of the potential and regularizations, with the data points sampled from the Gibbs density. The proof uses an elementary new derivation of the generalization error for random feature networks that does not apply the Rademacher or related complexities.
△ Less
Submitted 20 June, 2024;
originally announced June 2024.
-
Path integral molecular dynamics approximations of quantum canonical observables
Authors:
Xin Huang,
Petr Plechac,
Mattias Sandberg,
Anders Szepessy
Abstract:
Mean-field molecular dynamics based on path integrals is used to approximate canonical quantum observables for particle systems consisting of nuclei and electrons. A computational bottleneck is the sampling from the Gibbs density of the electron operator, which due to the fermion sign problem has a computational complexity that scales exponentially with the number of electrons. In this work we con…
▽ More
Mean-field molecular dynamics based on path integrals is used to approximate canonical quantum observables for particle systems consisting of nuclei and electrons. A computational bottleneck is the sampling from the Gibbs density of the electron operator, which due to the fermion sign problem has a computational complexity that scales exponentially with the number of electrons. In this work we construct an algorithm that approximates the mean-field Hamiltonian by path integrals for fermions. The algorithm is based on the determinant of a matrix with components based on Brownian bridges connecting permuted electron coordinates. The computational work for $n$ electrons is $\mathcal O(n^3)$, which reduces the computational complexity associated with the fermion sign problem. We analyze a bias resulting from this approximation and provide a computational error indicator. It remains to rigorously explain the surprisingly high accuracy.
△ Less
Submitted 28 November, 2023;
originally announced November 2023.
-
A Locally Corrected Multiblob Method with Hydrodynamically Matched Grids for the Stokes Mobility Problem
Authors:
Anna Broms,
Mattias Sandberg,
Anna-Karin Tornberg
Abstract:
Inexpensive numerical methods are key to enable simulations of systems of a large number of particles of different shapes in Stokes flow. Several approximate methods have been introduced for this purpose. We study the accuracy of the multiblob method for solving the Stokes mobility problem in free space, where the 3D geometry of a particle surface is discretised with spherical blobs and the pair-w…
▽ More
Inexpensive numerical methods are key to enable simulations of systems of a large number of particles of different shapes in Stokes flow. Several approximate methods have been introduced for this purpose. We study the accuracy of the multiblob method for solving the Stokes mobility problem in free space, where the 3D geometry of a particle surface is discretised with spherical blobs and the pair-wise interaction between blobs is described by the RPY-tensor. The paper aims to investigate and improve on the magnitude of the error in the solution velocities of the Stokes mobility problem using a combination of two different techniques: an optimally chosen grid of blobs and a pair-correction inspired by Stokesian dynamics. Optimisation strategies to determine a grid with a certain number of blobs are presented with the aim of matching the hydrodynamic response of a single accurately described ideal particle, alone in the fluid. Small errors in this self-interaction are essential as they determine the basic error level in a system of well-separated particles. With a good match, reasonable accuracy can be obtained even with coarse blob-resolutions of the particle surfaces. The error in the self-interaction is however sensitive to the exact choice of grid parameters and simply hand-picking a suitable blob geometry can lead to errors several orders of magnitude larger in size. The pair-correction is local and cheap to apply, and reduces on the error for more closely interacting particles. Two different types of geometries are considered: spheres and axisymmetric rods with smooth caps. The error in solutions to mobility problems is quantified for particles of varying inter-particle distances for systems containing a few particles, comparing to an accurate solution based on a second kind BIE-formulation where the quadrature error is controlled by employing quadrature by expansion (QBX).
△ Less
Submitted 22 July, 2022;
originally announced July 2022.
-
Canonical mean-field molecular dynamics derived from quantum mechanics
Authors:
Xin Huang,
Petr Plechac,
Mattias Sandberg,
Anders Szepessy
Abstract:
Canonical quantum correlation observables can be approximated by classical molecular dynamics. In the case of low temperature the ab initio molecular dynamics potential energy is based on the ground state electron eigenvalue problem and the accuracy has been proven to be $\mathcal O(M^{-1})$, provided the first electron eigenvalue gap is sufficiently large compared to the given temperature and…
▽ More
Canonical quantum correlation observables can be approximated by classical molecular dynamics. In the case of low temperature the ab initio molecular dynamics potential energy is based on the ground state electron eigenvalue problem and the accuracy has been proven to be $\mathcal O(M^{-1})$, provided the first electron eigenvalue gap is sufficiently large compared to the given temperature and $M$ is the ratio of nuclei and electron masses. For higher temperature eigenvalues corresponding to excited electron states are required to obtain $\mathcal O(M^{-1})$ accuracy and the derivations assume that all electron eigenvalues are separated, which for instance excludes conical intersections. This work studies a mean-field molecular dynamics approximation where the mean-field Hamiltonian for the nuclei is the partial trace $h:={\rm Tr}(H e^{-βH})/{\rm Tr}(e^{-βH})$ with respect to the electron degrees of freedom and $H$ is the Weyl symbol corresponding to a quantum many body Hamiltonian $\widehat{H}$. It is proved that the mean-field molecular dynamics approximates canonical quantum correlation observables with accuracy $\mathcal O (M^{-1}+ tε^2)$, for correlation time $t$ where $ε^2$ is related to the variance of mean value approximation $h$. Furthermore, the proof derives a precise asymptotic representation of the Weyl symbol of the Gibbs density operator using a path integral formulation. Numerical experiments on a model problem with one nuclei and two electron states show that the mean-field dynamics has similar or better accuracy than standard molecular dynamics based on the ground state electron eigenvalue.
△ Less
Submitted 23 January, 2023; v1 submitted 22 November, 2021;
originally announced November 2021.
-
Smaller generalization error derived for a deep residual neural network compared to shallow networks
Authors:
Aku Kammonen,
Jonas Kiessling,
Petr Plecháč,
Mattias Sandberg,
Anders Szepessy,
Raúl Tempone
Abstract:
Estimates of the generalization error are proved for a residual neural network with $L$ random Fourier features layers
$\bar z_{\ell+1}=\bar z_\ell + \mathrm{Re}\sum_{k=1}^K\bar b_{\ell k}e^{\mathrm{i}ω_{\ell k}\bar z_\ell}+ \mathrm{Re}\sum_{k=1}^K\bar c_{\ell k}e^{\mathrm{i}ω'_{\ell k}\cdot x}$. An optimal distribution for the frequencies $(ω_{\ell k},ω'_{\ell k})$ of the random Fourier feature…
▽ More
Estimates of the generalization error are proved for a residual neural network with $L$ random Fourier features layers
$\bar z_{\ell+1}=\bar z_\ell + \mathrm{Re}\sum_{k=1}^K\bar b_{\ell k}e^{\mathrm{i}ω_{\ell k}\bar z_\ell}+ \mathrm{Re}\sum_{k=1}^K\bar c_{\ell k}e^{\mathrm{i}ω'_{\ell k}\cdot x}$. An optimal distribution for the frequencies $(ω_{\ell k},ω'_{\ell k})$ of the random Fourier features $e^{\mathrm{i}ω_{\ell k}\bar z_\ell}$ and $e^{\mathrm{i}ω'_{\ell k}\cdot x}$ is derived. This derivation is based on the corresponding generalization error for the approximation of the function values $f(x)$. The generalization error turns out to be smaller than the estimate ${\|\hat f\|^2_{L^1(\mathbb{R}^d)}}/{(KL)}$ of the generalization error for random Fourier features with one hidden layer and the same total number of nodes $KL$, in the case the $L^\infty$-norm of $f$ is much less than the $L^1$-norm of its Fourier transform $\hat f$. This understanding of an optimal distribution for random features is used to construct a new training method for a deep residual network. Promising performance of the proposed new algorithm is demonstrated in computational experiments.
△ Less
Submitted 14 April, 2021; v1 submitted 5 October, 2020;
originally announced October 2020.
-
Adaptive random Fourier features with Metropolis sampling
Authors:
Aku Kammonen,
Jonas Kiessling,
Petr Plecháč,
Mattias Sandberg,
Anders Szepessy
Abstract:
The supervised learning problem to determine a neural network approximation $\mathbb{R}^d\ni x\mapsto\sum_{k=1}^K\hatβ_k e^{\mathrm{i}ω_k\cdot x}$ with one hidden layer is studied as a random Fourier features algorithm. The Fourier features, i.e., the frequencies $ω_k\in\mathbb{R}^d$, are sampled using an adaptive Metropolis sampler. The Metropolis test accepts proposal frequencies $ω_k'$, having…
▽ More
The supervised learning problem to determine a neural network approximation $\mathbb{R}^d\ni x\mapsto\sum_{k=1}^K\hatβ_k e^{\mathrm{i}ω_k\cdot x}$ with one hidden layer is studied as a random Fourier features algorithm. The Fourier features, i.e., the frequencies $ω_k\in\mathbb{R}^d$, are sampled using an adaptive Metropolis sampler. The Metropolis test accepts proposal frequencies $ω_k'$, having corresponding amplitudes $\hatβ_k'$, with the probability $\min\big\{1, (|\hatβ_k'|/|\hatβ_k|)^γ\big\}$, for a certain positive parameter $γ$, determined by minimizing the approximation error for given computational work. This adaptive, non-parametric stochastic method leads asymptotically, as $K\to\infty$, to equidistributed amplitudes $|\hatβ_k|$, analogous to deterministic adaptive algorithms for differential equations. The equidistributed amplitudes are shown to asymptotically correspond to the optimal density for independent samples in random Fourier features methods. Numerical evidence is provided in order to demonstrate the approximation properties and efficiency of the proposed algorithm. The algorithm is tested both on synthetic data and a real-world high-dimensional benchmark.
△ Less
Submitted 26 November, 2020; v1 submitted 21 July, 2020;
originally announced July 2020.
-
Computable error estimates for finite element approximations of elliptic partial differential equations with rough stochastic data
Authors:
Eric Joseph Hall,
Håkon Hoel,
Mattias Sandberg,
Anders Szepessy,
Raúl Tempone
Abstract:
We derive computable error estimates for finite element approximations of linear elliptic partial differential equations (PDE) with rough stochastic coefficients. In this setting, the exact solutions contain high frequency content that standard a posteriori error estimates fail to capture. We propose goal-oriented estimates, based on local error indicators, for the pathwise Galerkin and expected q…
▽ More
We derive computable error estimates for finite element approximations of linear elliptic partial differential equations (PDE) with rough stochastic coefficients. In this setting, the exact solutions contain high frequency content that standard a posteriori error estimates fail to capture. We propose goal-oriented estimates, based on local error indicators, for the pathwise Galerkin and expected quadrature errors committed in standard, continuous, piecewise linear finite element approximations. Derived using easily validated assumptions, these novel estimates can be computed at a relatively low cost and have applications to subsurface flow problems in geophysics where the conductivities are assumed to have lognormal distributions with low regularity. Our theory is supported by numerical experiments on test problems in one and two dimensions.
△ Less
Submitted 26 August, 2016; v1 submitted 9 October, 2015;
originally announced October 2015.
-
An adaptive mass algorithm for Car-Parrinello and Ehrenfest ab initio molecular dynamics
Authors:
Ashraful Kadir,
Mattias Sandberg,
Anders Szepessy
Abstract:
Ehrenfest and Car-Parrinello molecular dynamics are computational alternatives to approximate Born-Oppenheimer molecular dynamics without solving the electron eigenvalue problem at each time-step. A non-trivial issue is to choose the artificial electron mass parameter appearing in the Car-Parrinello method to achieve both good accuracy and high computational efficiency. In this paper, we propose a…
▽ More
Ehrenfest and Car-Parrinello molecular dynamics are computational alternatives to approximate Born-Oppenheimer molecular dynamics without solving the electron eigenvalue problem at each time-step. A non-trivial issue is to choose the artificial electron mass parameter appearing in the Car-Parrinello method to achieve both good accuracy and high computational efficiency. In this paper, we propose an algorithm, motivated by the Landau-Zener probability, to systematically choose an artificial mass dynamically, which makes the Car-Parrinello and Ehrenfest molecular dynamics methods dependent only on the problem data. Numerical experiments for simple model problems show that the time-dependent adaptive artificial mass parameter improves the efficiency of the Car-Parrinello and Ehrenfest molecular dynamics.
△ Less
Submitted 17 September, 2014;
originally announced September 2014.
-
An a posteriori error estimate for Symplectic Euler approximation of optimal control problems
Authors:
Jesper Karlsson,
Stig Larsson,
Mattias Sandberg,
Anders Szepessy,
Raùl Tempone
Abstract:
This work focuses on numerical solutions of optimal control problems. A time discretization error representation is derived for the approximation of the associated value function. It concerns Symplectic Euler solutions of the Hamiltonian system connected with the optimal control problem. The error representation has a leading order term consisting of an error density that is computable from Symple…
▽ More
This work focuses on numerical solutions of optimal control problems. A time discretization error representation is derived for the approximation of the associated value function. It concerns Symplectic Euler solutions of the Hamiltonian system connected with the optimal control problem. The error representation has a leading order term consisting of an error density that is computable from Symplectic Euler solutions. Under an assumption of the pathwise convergence of the approximate dual function as the maximum time step goes to zero, we prove that the remainder is of higher order than the leading error density part in the error representation. With the error representation, it is possible to perform adaptive time step**. We apply an adaptive algorithm originally developed for ordinary differential equations. The performance is illustrated by numerical tests.
△ Less
Submitted 31 July, 2014;
originally announced July 2014.
-
Computational error estimates for Born-Oppenheimer molecular dynamics with nearly crossing potential surfaces
Authors:
Christian Bayer,
Hakon Hoel,
Ashraful Kadir,
Petr Plechac,
Mattias Sandberg,
Anders Szepessy
Abstract:
The difference of the values of observables for the time-independent Schroedinger equation, with matrix valued potentials, and the values of observables for ab initio Born-Oppenheimer molecular dynamics, of the ground state, depends on the probability to be in excited states and the electron/nuclei mass ratio. The paper first proves an error estimate (depending on the electron/nuclei mass ratio an…
▽ More
The difference of the values of observables for the time-independent Schroedinger equation, with matrix valued potentials, and the values of observables for ab initio Born-Oppenheimer molecular dynamics, of the ground state, depends on the probability to be in excited states and the electron/nuclei mass ratio. The paper first proves an error estimate (depending on the electron/nuclei mass ratio and the probability to be in excited states) for this difference of microcanonical observables, assuming that molecular dynamics space-time averages converge, with a rate related to the maximal Lyapunov exponent. The error estimate is uniform in the number of particles and the analysis does not assume a uniform lower bound on the spectral gap of the electron operator and consequently the probability to be in excited states can be large. A numerical method to determine the probability to be in excited states is then presented, based on Ehrenfest molecular dynamics and stability analysis of a perturbed eigenvalue problem.
△ Less
Submitted 12 May, 2015; v1 submitted 14 May, 2013;
originally announced May 2013.
-
The Forward Euler Scheme for Nonconvex Lipschitz Differential Inclusions Converges with Rate One
Authors:
Mattias Sandberg
Abstract:
In a previous paper it was shown that the Forward Euler method applied to differential inclusions where the right-hand side is a Lipschitz continuous set-valued function with uniformly bounded, compact values, converges with rate one. The convergence, which was there in the sense of reachable sets, is in this paper strengthened to the sense of convergence of solution paths. An improvement of the…
▽ More
In a previous paper it was shown that the Forward Euler method applied to differential inclusions where the right-hand side is a Lipschitz continuous set-valued function with uniformly bounded, compact values, converges with rate one. The convergence, which was there in the sense of reachable sets, is in this paper strengthened to the sense of convergence of solution paths. An improvement of the error constant is given for the case when the set-valued function consists of a small number of smooth ordinary functions.
△ Less
Submitted 29 January, 2009;
originally announced January 2009.
-
Extended Applicability of the Symplectic Pontryagin Method
Authors:
Mattias Sandberg
Abstract:
The Symplectic Pontryagin method was introduced in a previous paper. This work shows that this method is applicable under less restrictive assumptions. Existence of solutions to the Symplectic Pontryagin scheme are shown to exist without the previous assumption on a bounded gradient of the discrete dual variable. The convergence proof uses the representation of solutions to a Hamilton-Jacobi-Bel…
▽ More
The Symplectic Pontryagin method was introduced in a previous paper. This work shows that this method is applicable under less restrictive assumptions. Existence of solutions to the Symplectic Pontryagin scheme are shown to exist without the previous assumption on a bounded gradient of the discrete dual variable. The convergence proof uses the representation of solutions to a Hamilton-Jacobi-Bellman equation as the value function of an associated variation problem.
△ Less
Submitted 29 January, 2009;
originally announced January 2009.
-
Convergence rates for an optimally controlled Ginzburg-Landau equation
Authors:
Mattias Sandberg
Abstract:
An optimal control problem related to the probability of transition between stable states for a thermally driven Ginzburg-Landau equation is considered. The value function for the optimal control problem with a spatial discretization is shown to converge quadratically to the value function for the original problem. This is done by using that the value functions solve similar Hamilton-Jacobi equa…
▽ More
An optimal control problem related to the probability of transition between stable states for a thermally driven Ginzburg-Landau equation is considered. The value function for the optimal control problem with a spatial discretization is shown to converge quadratically to the value function for the original problem. This is done by using that the value functions solve similar Hamilton-Jacobi equations, the equation for the original problem being defined on an infinite dimensional Hilbert space. Time discretization is performed using the Symplectic Euler method. Imposing a reasonable condition this method is shown to be convergent of order one in time, with a constant independent of the spatial discretization.
△ Less
Submitted 10 September, 2008;
originally announced September 2008.
