Jacobi eigenvalue algorithm

E697940

The Jacobi eigenvalue algorithm is an iterative numerical method for computing all eigenvalues and eigenvectors of a real symmetric matrix by applying a sequence of orthogonal similarity transformations.

Jump to: Statements Referenced by

Statements (48)

Predicate Object
instanceOf eigenvalue algorithm
iterative method
matrix diagonalization method
numerical algorithm
accuracy high relative accuracy for well-separated eigenvalues
advantage conceptually simple
high accuracy for eigenvectors
produces orthogonal eigenvectors explicitly
application principal component analysis
quantum mechanics eigenproblems
vibration analysis
complexity O(n^3) for an n-by-n matrix
computes eigenvalues
eigenvectors
convergenceType global convergence for symmetric matrices
coreOperation Jacobi rotations NERFINISHED
plane rotations
disadvantage not optimal for large-scale problems
relatively slow compared to modern methods
field numerical linear algebra
goal diagonalize a symmetric matrix
historicalPeriod 19th century origin
implementation used in some LAPACK routines historically
lessSuitableFor sparse matrices
very large matrices
matrixClass normal matrices (via unitary version)
methodType iterative diagonalization
namedAfter Carl Gustav Jacob Jacobi NERFINISHED
operatesOn Hermitian matrices
real symmetric matrices
output diagonal matrix of eigenvalues
orthogonal matrix of eigenvectors
pivotSelection largest off-diagonal element strategy
propertyPreserved eigenvalues
orthogonality of eigenvectors
symmetry of the matrix
relatedTo Householder transformation methods NERFINISHED
QR algorithm NERFINISHED
power iteration
requires selection of pivot elements
stability numerically stable for symmetric problems
stoppingCriterion small off-diagonal elements
suitableFor small to medium size dense matrices
transformationType orthogonal similarity transformations
uses Givens rotations in some formulations
variant blocked Jacobi method
cyclic Jacobi method NERFINISHED
parallel Jacobi method

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.

Jacobi matrix usedIn Jacobi eigenvalue algorithm