numpy.linalg.eig#
- linalg.eig(a)[source]#
Compute the eigenvalues and right eigenvectors of a square array.
- Parameters:
- a(…, M, M) array
Matrices for which the eigenvalues and right eigenvectors willbe computed
- Returns:
- A namedtuple with the following attributes:
- eigenvalues(…, M) array
The eigenvalues, each repeated according to its multiplicity.The eigenvalues are not necessarily ordered. The resultingarray will be of complex type, unless the imaginary part iszero in which case it will be cast to a real type. Whenais real the resulting eigenvalues will be real (0 imaginarypart) or occur in conjugate pairs
- eigenvectors(…, M, M) array
The normalized (unit “length”) eigenvectors, such that thecolumn
eigenvectors[:,i]is the eigenvector corresponding to theeigenvalueeigenvalues[i].
- Raises:
- LinAlgError
If the eigenvalue computation does not converge.
See also
eigvalseigenvalues of a non-symmetric array.
eigheigenvalues and eigenvectors of a real symmetric or complex Hermitian (conjugate symmetric) array.
eigvalsheigenvalues of a real symmetric or complex Hermitian (conjugate symmetric) array.
scipy.linalg.eigSimilar function in SciPy that also solves the generalized eigenvalue problem.
scipy.linalg.schurBest choice for unitary and other non-Hermitian normal matrices.
Notes
Broadcasting rules apply, see the
numpy.linalgdocumentation fordetails.This is implemented using the
_geevLAPACK routines which computethe eigenvalues and eigenvectors of general square arrays.The numberw is an eigenvalue ofa if there exists a vectorv suchthat
a@v=w*v. Thus, the arraysa,eigenvalues, andeigenvectors satisfy the equationsa@eigenvectors[:,i]=eigenvalues[i]*eigenvectors[:,i]for\(i \in \{0,...,M-1\}\).The arrayeigenvectors may not be of maximum rank, that is, some of thecolumns may be linearly dependent, although round-off error may obscurethat fact. If the eigenvalues are all different, then theoretically theeigenvectors are linearly independent anda can be diagonalized by asimilarity transformation usingeigenvectors, i.e,
inv(eigenvectors)@a@eigenvectorsis diagonal.For non-Hermitian normal matrices the SciPy function
scipy.linalg.schuris preferred because the matrixeigenvectors is guaranteed to beunitary, which is not the case when usingeig. The Schur factorizationproduces an upper triangular matrix rather than a diagonal matrix, but fornormal matrices only the diagonal of the upper triangular matrix isneeded, the rest is roundoff error.Finally, it is emphasized thateigenvectors consists of theright (asin right-hand side) eigenvectors ofa. A vectory satisfying
y.T@a=z*y.Tfor some numberz is called aleft eigenvector ofa,and, in general, the left and right eigenvectors of a matrix are notnecessarily the (perhaps conjugate) transposes of each other.References
G. Strang,Linear Algebra and Its Applications, 2nd Ed., Orlando, FL,Academic Press, Inc., 1980, Various pp.
Examples
>>>importnumpyasnp>>>fromnumpyimportlinalgasLA
(Almost) trivial example with real eigenvalues and eigenvectors.
>>>eigenvalues,eigenvectors=LA.eig(np.diag((1,2,3)))>>>eigenvaluesarray([1., 2., 3.])>>>eigenvectorsarray([[1., 0., 0.], [0., 1., 0.], [0., 0., 1.]])
Real matrix possessing complex eigenvalues and eigenvectors;note that the eigenvalues are complex conjugates of each other.
>>>eigenvalues,eigenvectors=LA.eig(np.array([[1,-1],[1,1]]))>>>eigenvaluesarray([1.+1.j, 1.-1.j])>>>eigenvectorsarray([[0.70710678+0.j , 0.70710678-0.j ], [0. -0.70710678j, 0. +0.70710678j]])
Complex-valued matrix with real eigenvalues (but complex-valuedeigenvectors); note that
a.conj().T==a, i.e.,a is Hermitian.>>>a=np.array([[1,1j],[-1j,1]])>>>eigenvalues,eigenvectors=LA.eig(a)>>>eigenvaluesarray([2.+0.j, 0.+0.j])>>>eigenvectorsarray([[ 0. +0.70710678j, 0.70710678+0.j ], # may vary [ 0.70710678+0.j , -0. +0.70710678j]])
Be careful about round-off error!
>>>a=np.array([[1+1e-9,0],[0,1-1e-9]])>>># Theor. eigenvalues are 1 +/- 1e-9>>>eigenvalues,eigenvectors=LA.eig(a)>>>eigenvaluesarray([1., 1.])>>>eigenvectorsarray([[1., 0.], [0., 1.]])