Yunkai Zhou

On the decay rate of Hankel singular values and related issues

This paper investigates the decay rate of the Hankel singular values of linear dynamical systems. This issue is of considerable interest in model reduction by means of balanced truncation, for instance, since the sum of the neglected singular values provides an upper bound for an appropriate norm of the approximation error. The decay rate involves a new set of invariants associated with a linear system, which are obtained by evaluating a modified transfer function at the poles of the system. These considerations are equivalent to studying the decay rate of the eigenvalues of the product of the solutions of two Lyapunov equations. The related problem of determining the decay rate of the eigenvalues of the solution to one Lyapunov equation will also be addressed. Very often these eigenvalues, like the Hankel singular values, are rapidly decaying. This fact has motivated the development of several algorithms for computing low-rank approximate solutions to Lyapunov equations. However, until now, conditions assuring rapid decay have not been well understood. Such conditions are derived here by relating the solution to a numerically low-rank Cauchy matrix determined by the poles of the system. Bounds explaining rapid decay rates are obtained under some mild conditions.

Self-consistent-field calculations using Chebyshev-filtered subspace iteration

The power of density functional theory is often limited by the high computational demand in solving an eigenvalue problem at each self-consistent-field (SCF) iteration. The method presented in this paper replaces the explicit eigenvalue calculations by an approximation of the wanted invariant subspace, obtained with the help of well-selected Chebyshev polynomial filters. In this approach, only the initial SCF iteration requires solving an eigenvalue problem, in order to provide a good initial subspace. In the remaining SCF iterations, no iterative eigensolvers are involved. Instead, Chebyshev polynomials are used to refine the subspace. The subspace iteration at each step is easily five to ten times faster than solving a corresponding eigenproblem by the most efficient eigen-algorithms. Moreover, the subspace iteration reaches self-consistency within roughly the same number of steps as an eigensolver-based approach. This results in a significantly faster SCF iteration.

Parallel self-consistent-field calculations via Chebyshev-filtered subspace acceleration.

Solving the Kohn-Sham eigenvalue problem constitutes the most computationally expensive part in self-consistent density functional theory (DFT) calculations. In a previous paper, we have proposed a nonlinear Chebyshev-filtered subspace iteration method, which avoids computing explicit eigenvectors except at the first self-consistent-field (SCF) iteration. The method may be viewed as an approach to solve the original nonlinear Kohn-Sham equation by a nonlinear subspace iteration technique, without emphasizing the intermediate linearized Kohn-Sham eigenvalue problems. It reaches self-consistency within a similar number of SCF iterations as eigensolver-based approaches. However, replacing the standard diagonalization at each SCF iteration by a Chebyshev subspace filtering step results in a significant speedup over methods based on standard diagonalization. Here, we discuss an approach for implementing this method in multi-processor, parallel environment. Numerical results are presented to show that the method enables to perform a class of highly challenging DFT calculations that were not feasible before.

Direct methods for matrix Sylvester and Lyapunov equations

We revisit the two standard dense methods for matrix Sylvester and Lyapunov equations: the Bartels-Stewart method for A1X

Evolution of Magnetism in Iron from the Atom to the Bulk

The evolution of the magnetic moment in iron clusters containing 20-400 atoms is investigated using first-principles numerical calculations based on density-functional theory and real-space pseudopotentials. Three families of clusters are studied, characterized by the arrangement of atoms: icosahedral, body-centered cubic centered on an atom site, and body-centered cubic centered on the bridge between two neighboring atoms. We find an overall decrease of magnetic moment as the clusters grow in size towards the bulk limit. Clusters with faceted surfaces are predicted to have magnetic moment lower than other clusters with similar size. As a result, the magnetic moment is observed to decrease as function of size in a nonmonotonic manner, which explains measurements performed at low temperatures.

A Chebyshev-Davidson Algorithm for Large Symmetric Eigenproblems

A polynomial filtered Davidson-type algorithm is proposed for symmetric eigenproblems, in which the correction-equation of the Davidson approach is replaced by a polynomial filtering step. The new approach has better global convergence and robustness properties when compared with standard Davidson-type methods. The typical filter used in this paper is based on Chebyshev polynomials. The goal of the polynomial filter is to amplify components of the desired eigenvectors in the subspace, which has the effect of reducing both the number of steps required for convergence and the cost in orthogonalizations and restarts. Numerical results are presented to show the effectiveness of the proposed approach.

Block Krylov-Schur Method for Large Symmetric Eigenvalue Problems ∗

Stewart’s Krylov–Schur algorithm offers two advantages over Sorensen’s implicitly restarted Arnoldi (IRA) algorithm. The first is ease of deflation of converged Ritz vectors, the second is the avoidance of the potential forward instability of the QR algorithm. In this paper we develop a block version of the Krylov–Schur algorithm for symmetric eigenproblems. Details of this block algorithm are discussed, including how to handle rank deficient cases and how to use varying block sizes. Numerical results on the efficiency of the block Krylov–Schur method are reported.

Dynamic Multiple-Fault Diagnosis With Imperfect Tests

Fault diagnosis is the process of identifying the failure sources of a malfunctioning system by observing their effects at various test points. It has a number of applications in engineering and medicine. In this paper, we present a near-optimal algorithm for dynamic multiple fault diagnosis in complex systems. This problem involves on-board diagnosis of the most likely set of faults and their time-evolution based on blocks of unreliable test outcomes over time. The dynamic multiple fault diagnosis (dMFD) problem is an intractable NP-hard combinatorial optimization problem. Consequently, we decompose the dMFD problem into a series of decoupled sub-problems, and develop a successive Lagrangian relaxation algorithm (SLRA) with backtracking to obtain a near-optimal solution for the problem. SLRA solves the sub-problems at each sample point by a Lagrangian relaxation method, and shares Lagrange multipliers at successive time points to speed up convergence. In addition, we apply a backtracking technique to further maximize the likelihood of obtaining the most likely evolution of failure sources and to minimize the effects of imperfect tests.

Studies on Jacobi-Davidson, Rayleigh quotient iteration, inverse iteration generalized Davidson and Newton updates

SUMMARY We study Davidson-type subspace eigensolvers. Correction equations of Jacobi–Davidson and several other schemes are reviewed. New correction equations are derived. A general correction equation is constructed, existing correction equations may be considered as special cases of this general equation. The main theme of this study is to identify the essential common ingredient that leads to the efficiency of a diverse form of Davidson-type methods. We emphasize the importance of the approximate Rayleigh-quotient-iteration direction. Copyright q 2006 John Wiley & Sons, Ltd.

Chebyshev-filtered subspace iteration method free of sparse diagonalization for solving the Kohn-Sham equation

First-principles density functional theory (DFT) calculations for the electronic structure problem require a solution of the Kohn-Sham equation, which requires one to solve a nonlinear eigenvalue problem. Solving the eigenvalue problem is usually the most expensive part in DFT calculations. Sparse iterative diagonalization methods that compute explicit eigenvectors can quickly become prohibitive for large scale problems. The Chebyshev-filtered subspace iteration (CheFSI) method avoids most of the explicit computation of eigenvectors and results in a significant speedup over iterative diagonalization methods for the DFT self-consistent field (SCF) calculations. However, the original formulation of the CheFSI method utilizes a sparse iterative diagonalization at the first SCF step to provide initial vectors for subspace filtering at latter SCF steps. This diagonalization is expensive for large scale problems. We develop a new initial filtering step to avoid completely this diagonalization, thus making the CheFSI method free of sparse iterative diagonalizations at all SCF steps. Our new approach saves memory usage and can be two to three times faster than the original CheFSI method.