A matrix norm on square matrices over or
, in the submultiplicative convention
used here, assigns a nonnegative number
to each matrix
and has the properties
1. when
and
iff
,
2. for any scalar
,
3. ,
4. .
Some authors require only the first three properties and call the fourth property submultiplicativity (Higham 2021). Norms satisfying the fourth property obey the
following eigenvalue bound. If is a nonsingular matrix
and
is any of its eigenvalues,
then
|
(1)
|
The induced matrix -norm
is also defined for an
matrix
, with
,
and
, by
|
(2)
|
where
is a vector norm. The cases
,
2, and
have the explicit formulas below. For other
, an explicit formula is not generally
available (Higham 1992, pp. 539-540).
Matrix norms are implemented as Norm[m, p], where
may be 1, 2, Infinity, or "Frobenius".
The maximum absolute column sum norm is defined as
|
(3)
|
The spectral norm ,
which is the square root of the maximum eigenvalue
of
(where
is the conjugate transpose),
|
(4)
|
is often referred to as "the" matrix norm.
The maximum absolute row sum norm is defined by
|
(5)
|
,
, and
satisfy the inequality
|
(6)
|