Operator norm

Norm of a linear map defined by its maximal expansion of unit vectors.
Operator norm

An operator norm of a T:VWT:V\to W between (V,V)(V,\|\cdot\|_V) and (W,W)(W,\|\cdot\|_W) is the quantity

T=supv0T(v)WvV=supvV=1T(v)W, \|T\|=\sup_{v\ne 0}\frac{\|T(v)\|_W}{\|v\|_V}=\sup_{\|v\|_V=1}\|T(v)\|_W,

with the understanding that the supremum may be ++\infty in general. When T<\|T\|<\infty, one says TT is bounded.

For linear maps between normed spaces, finiteness of the operator norm is equivalent to being . The operator norm makes the collection of bounded linear maps into a normed vector space and specializes to a norm on when V=WV=W.

Examples:

  • If T(v)=cvT(v)=c\,v on a normed space, then T=c\|T\|=|c|.
  • For the projection P(x,y)=(x,0)P(x,y)=(x,0) on R2\mathbb{R}^2 with the Euclidean norm, P=1\|P\|=1.
  • For a diagonal matrix A=diag(d1,,dn)A=\operatorname{diag}(d_1,\dots,d_n) acting on Rn\mathbb{R}^n with the max norm, the induced operator norm is A=maxidi\|A\|=\max_i |d_i|.