Backward error of an approximate eigenpair Created 2026-09-24 Updated 2026-11-05
Let be an eigenpair residual, with . Among unrestricted matrix perturbations , the smallest spectral norm making an exact eigenpair of is .
Indeed, exactness requires , so the induced matrix norm inequality gives . Equality is attained by the rank-one matrix
where is conjugate transpose. Its spectral norm is . For real data, use ordinary transpose instead.
This is an unstructured backward error: requiring to preserve a symmetric matrix structure or a sparsity pattern is an additional constraint. A small backward error also need not give a comparably small eigenvalue error for a nonnormal matrix.
Linear solver Created 2026-10-25 Updated 2026-11-05
A linear solver is an algorithm that computes a solution or controlled approximation to a linear system . For an invertible matrix, the mathematical solution is unique. A direct linear solver usually computes a matrix factorization, such as Pivoted LU decomposition, followed by triangular solves. An iterative method instead updates an approximate vector, often using matrix-vector products and a matrix preconditioner.
A stopping rule should reflect the required accuracy. The linear-system residual can be scaled as
when the denominator is positive, using compatible vector norms and an induced matrix norm. This compares equation error with the input scale; a condition number is still needed to interpret it as a forward error guarantee. In a differential equation computation, the linear solver tolerance must be small enough relative to the intended discretization accuracy. A stable time step alone does not control errors left by an inaccurate inner solve.
For a real symmetric positive-definite matrix , the energy norm is . In the defining supremum of the induced matrix norm, substitute . The matrix square root is an invertible matrix, so this ranges over every nonzero vector and proves the displayed identity. This matrix similarity is useful because contraction in the energy norm becomes an ordinary Euclidean operator norm estimate after a change of coordinates.
A real square orthogonal matrix satisfies , so it preserves the Euclidean norm. The columns of are orthonormal, hence . The induced matrix norm therefore gives
For the left truncation, , and equality is guaranteed for arbitrary only when . Indeed, if , take , where is an omitted column of . Then but . Merely taking does not guarantee preservation for an arbitrary .
A full QR decomposition of a real matrix is , with orthogonal and upper trapezoidal; for a square matrix, is upper triangular. Fix the following sign convention for a Givens rotation: it is the identity except for its block
The block transpose times the block is because , proving orthogonality. Writing , the choice gives and . To zero the other coordinate take , giving . If , both entries are already zero and the identity suffices.
For a tridiagonal matrix, successively rotate adjacent row pairs , each time zeroing the current entry . Earlier columns remain triangular because both affected rows are already zero there. The result is upper triangular, so gives its QR decomposition. Only rotations are required; the resulting has at most two superdiagonals.
For a full square matrix, process columns . Within column , for , rotate rows to eliminate the current entry. Previously cleared columns stay clear. With in application order, , one has . This is QR decomposition by Givens rotations.
The vector Euclidean norm is , the ordinary length from the origin. The induced matrix norm is
Geometrically this is the largest stretch of a unit vector. A linear map sends the unit sphere to an ellipsoid, possibly degenerate; its largest semiaxis has length . Equivalently the matrix 2-norm is the largest singular value.
The induced matrix norm associated with the given energy norm is a supremum of length ratios. The invertible change of variables gives
The last expression is the Euclidean operator norm. This is the matrix norm induced by an energy norm: transforming to coordinates with Euclidean lengths transforms the operator by the corresponding matrix similarity.
Row-column bound for the matrix 2-norm Created 2026-10-31 Updated 2026-11-05
For each row, the weighted Cauchy-Schwarz inequality gives . Summing over rows bounds the result by the maximum absolute row sum times the maximum absolute column sum times . Taking square roots and the supremum proves the displayed induced matrix norm inequality. It gives a Euclidean estimate directly from entries without computing singular values.