Convergence analysis of adaptive DIIS algorithms with application to electronic ground state calculations
From MaRDI portal
Publication:5034831
DOI10.1051/M2AN/2021069zbMATH Open1492.65008arXiv2002.12850OpenAlexW3216904700MaRDI QIDQ5034831FDOQ5034831
Authors: Maxime Chupin, Mi-Song Dupuy, Guillaume Legendre, Éric Séré 
Publication date: 21 February 2022
Published in: ESAIM: Mathematical Modelling and Numerical Analysis (Search for Journal in Brave)
Abstract: This paper deals with a general class of algorithms for the solution of fixed-point problems that we refer to as emph{Anderson--Pulay acceleration}. This family includes the DIIS technique and its variant sometimes called commutator-DIIS, both introduced by Pulay in the 1980s to accelerate the convergence of self-consistent field procedures in quantum chemistry, as well as the related Anderson acceleration which dates back to the 1960s, and the wealth of techniques they have inspired. Such methods aim at accelerating the convergence of any fixed-point iteration method by combining several iterates in order to generate the next one at each step. This extrapolation process is characterised by its emph{depth}, i.e. the number of previous iterates stored, which is a crucial parameter for the efficiency of the method. It is generally fixed to an empirical value. In the present work, we consider two parameter-driven mechanisms to let the depth vary along the iterations. In the first one, the depth grows until a certain nondegeneracy condition is no longer satisfied; then the stored iterates (save for the last one) are discarded and the method "restarts". In the second one, we adapt the depth continuously by eliminating at each step some of the oldest, less relevant, iterates. In an abstract and general setting, we prove under natural assumptions the local convergence and acceleration of these two adaptive Anderson--Pulay methods, and we show that one can theoretically achieve a superlinear convergence rate with each of them. We then investigate their behaviour in quantum chemistry calculations. These numerical experiments show that both adaptive variants exhibit a faster convergence than a standard fixed-depth scheme, and require on average less computational effort per iteration. This study is complemented by a review of known facts on the DIIS, in particular its link with the Anderson acceleration and some multisecant-type quasi-Newton methods.
Full work available at URL: https://arxiv.org/abs/2002.12850
Recommendations
- An analysis for the DIIS acceleration method used in quantum chemistry calculations
- On one-point iterations and DIIS
- Convergence of the EDIIS algorithm for nonlinear equations
- Reformulation of the Anderson method using singular value decomposition for stable convergence in self-consistent calculations
- Anderson acceleration for fixed-point iterations
Methods of quasi-Newton type (90C53) Numerical computation of solutions to systems of equations (65H10) Computational methods for problems pertaining to quantum theory (81-08) Applications to the sciences (65Z05) Acceleration of convergence in numerical analysis (65B99)
Cites Work
- Design and Application of a Gradient-Weighted Moving Finite Element Code I: in One Dimension
- Title not available (Why is that?)
- Damped Anderson Acceleration With Restarts and Monotonicity Control for Accelerating EM and EM-like Algorithms
- A comparative study on methods for convergence acceleration of iterative vector sequences
- Two classes of multisecant methods for nonlinear acceleration
- Anderson acceleration of the alternating projections method for computing the nearest correlation matrix
- Anderson acceleration for fixed-point iterations
- GMRES: A Generalized Minimal Residual Algorithm for Solving Nonsymmetric Linear Systems
- Iterative Procedures for Nonlinear Integral Equations
- An analysis for the DIIS acceleration method used in quantum chemistry calculations
- A comprehensive introduction to differential geometry. Vol. 1-5
- Truncation Strategies for Optimal Krylov Subspace Methods
- A Class of Methods for Solving Nonlinear Simultaneous Equations
- Any Nonincreasing Convergence Curve is Possible for GMRES
- Existence of minimizers for Kohn-Sham models in quantum chemistry
- A characterization of the behavior of the Anderson acceleration on linear problems
- Convergence analysis for Anderson acceleration
- Krylov subspace acceleration for nonlinear multigrid schemes
- Accelerated block preconditioned gradient method for large scale wave functions calculations in density functional theory
- A new method for diagonalising large matrices
- On the similarities between the quasi-Newton inverse least squares method and GMRES
- New Developments in Molecular Orbital Theory
- On Q-order and R-order of convergence
- A nonlinear GMRES optimization algorithm for canonical tensor decomposition
- Title not available (Why is that?)
- On the convergence of SCF algorithms for the Hartree-Fock equations
- Converging self-consistent field equations in quantum chemistry – recent achievements and remaining challenges
- Nonlinear acceleration of coupled fluid-structure transient thermal problems by Anderson mixing
- Anderson acceleration and application to the three-temperature energy equations
- Leveraging Anderson acceleration for improved convergence of iterative solutions to transport systems
- The molecular orbital theory of chemical valency VIII. A method of calculating ionization potentials
- Nonlinear Krylov acceleration applied to a discrete ordinates formulation of the \(k\)-eigenvalue problem
- Shanks sequence transformations and Anderson acceleration
- A Proof That Anderson Acceleration Improves the Convergence Rate in Linearly Converging Fixed-Point Methods (But Not in Those Converging Quadratically)
- Broyden updating, the good and the bad!
- Comments on: ``Anderson acceleration, mixing and extrapolation
- Local improvement results for Anderson acceleration with inaccurate function evaluations
- Globally Convergent Type-I Anderson Acceleration for Nonsmooth Fixed-Point Iterations
- Convergence of the EDIIS algorithm for nonlinear equations
Cited In (12)
- Secant Acceleration of Sequential Residual Methods for Solving Large-Scale Nonlinear Systems of Equations
- A robust and efficient line search for self-consistent field iterations
- On one-point iterations and DIIS
- Convergence of a Constrained Vector Extrapolation Scheme
- Convergence of the EDIIS algorithm for nonlinear equations
- Descent Properties of an Anderson Accelerated Gradient Method with Restarting
- Anderson accelerating the preconditioned modulus approach for linear complementarity problems on second-order cones
- Converging self-consistent field equations in quantum chemistry – recent achievements and remaining challenges
- Body-ordered approximations of atomic properties
- Reformulation of the Anderson method using singular value decomposition for stable convergence in self-consistent calculations
- An analysis for the DIIS acceleration method used in quantum chemistry calculations
- nlTGCR: A Class of Nonlinear Acceleration Procedures Based on Conjugate Residuals
Uses Software
This page was built for publication: Convergence analysis of adaptive DIIS algorithms with application to electronic ground state calculations
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5034831)
