nLab
Hausdorff metric

The Hausdorff metric

Idea

The Hausdorff metric is a metric on the power set of a given metric space.

Definition

Let A be a metric space, regarded as a category enriched over V=[0,] (a Lawvere metric space). The enriched functor category [A,V] is, concretely, the set of short maps A[0,] with the supremum metric?, and the contravariant Yoneda embedding A op[A,V] sends a to d(a,).

Now, for any subset XA, each point xX gives rise to the representable functor d(x,), and we can define the functor d(X,):AV to be the coproduct of these representables over all xX. Concretely, this means

d(X,a)=inf xXd(x,a).d(X,a)=inf_{x\in X} d(x,a) .

Note that if X is closed, then it can be recovered from d(X,) as the set of points x such that d(X,x)=0. If X is not closed, then in this way we recover its closure.

Finally, since [A,V] is also a Lawvere metric space, we obtain an induced metric on the set of subspaces of A:

d(X,Y)=d(d(X,),d(Y,)).d(X,Y) = d(d(X,-), d(Y,-)) .

This metric is not symmetric (so a quasimetric); its symmetrization? is the Hausdorff metric. Equivalently, we could start out by considering instead the functor category [A,V sym] where V sym is the symmetrization of V=[0,].

References

  • Taking categories seriously, Reprints in Theory and Applications of Categories, No. 8, 2005, pp. 1–24. (pdf)
Revised on September 7, 2011 04:04:47 by Toby Bartels (75.88.82.16)