Danskin's theorem
From Wikipedia, the free encyclopedia
In convex analysis, Danskin's theorem is a theorem which provides information about the derivatives of a function of the form
The theorem has applications in optimization, where it sometimes is used to solve minimax problems.
[edit]Statement
The theorem applies to the following situation. Suppose is a continuous function of two arguments,
where is a compact set. Further assume that is convex in for every .
Under these conditions, Danskin's theorem provides conclusions regarding the differentiability of the function
To state these results, we define the set of maximizing points as
Danskin's theorem then provides the following results.
-
Convexity
- is convex. Directional derivatives
-
The
directional derivative of
in the direction
, denoted
, is given by
- where is the directional derivative of the function at in the direction . Derivative
-
is
differentiable at
if
consists of a single element
. In this case, the
derivative of
(or the
gradient of
if
is a vector) is given by
-
Subdifferential
-
If
is differentiable with respect to
for all
, and if
is continuous with respect to
for all
, then the
subdifferential of
is given by
- where indicates the convex hull operation.
[edit]References
- Bertsekas, Dimitri P. (1999). Nonlinear Programming. Belmont, MA: Athena Scientific. pp. 717. ISBN 1-886529-00-0.