next up previous
Next: About this document Up: Vector and Matrix Previous: Choice of Norm

Matrix Norms

Definition: is a matrix norm on matrices if it is a vector norm on an dimensional space:

  1. , and

Definition: Let

They are called mutually consistent if ,

Example: is the ``max norm".

, is the Frobenius norm.

Definition: Given , let be a vector norm on , be a vector norm on . Then

is called an operator norm or induced norm. The geometric interpretation of such a norm is that it is the maximum length of a unit vector after transformation by . Furthermore, an operator norm is a matrix norm (i.e. satisfies the property listed above).

Definition: A real square matrix is orthogonal, if . For an orthogonal matrix, all the rows and columns have and are orthogonal to one another.

Some properties of matrix and vector norms:

  1. for operator and (Frobenius norm + vector 2 norm)
  2. for operator and Froberiur norm.
  3. Max norm is not an operator norm.
  4. if are orthogonal for Frobenius and operator norm induced by .
  5. max absolute row sum.
  6. max absolute column sum.

  7. , where is the absolute value of the largest eigenvalue in magnitude of the matrix . (This is called the spectral radius of the matrix.) If is symmetric, corresponds to the maximum absolute eigenvalue, i.e.


Dinesh Manocha
Mon Jan 27 13:16:15 EST 1997