The Moreau-Yosida Regularization: Difference between revisions

From Optimal Transport Wiki
Jump to navigation Jump to search
(added a proposition (without proof) and a reference)
(added an example)
Line 12: Line 12:
:<math>g_k(x) := \inf\limits_{y \in X} \left[ g(y) + k d(x,y) \right].</math>
:<math>g_k(x) := \inf\limits_{y \in X} \left[ g(y) + k d(x,y) \right].</math>


If <math>k = 0</math>, then by definition <math>g_0</math> is constant and <math>g_0 \equiv \inf\limits_{y \in X} g(y)</math>.
 




==Examples==
==Examples==
(to be filled in, hopefully with pictures!)
* If <math>k = 0</math>, then by definition <math>g_0</math> is constant and <math>g_0 \equiv \inf\limits_{y \in X} g(y)</math>.
* If <math>g</math> is ''not'' proper, then <math>g_k = +\infty</math> for all <math>k \geq 0</math>.
 
Take <math>(X,d) := (\mathbb{R},|\cdot|)</math>. If <math>g</math> is finite-valued and differentiable, we can explicitly write down <math>g_k</math>. Then for a fixed <math>x \in \mathbb{R}</math>, the map <math>g_{k,x} : y \mapsto y^2 + k|x - y|</math> is continuous everywhere and differentiable everywhere except for when <math>y = x</math>, where the derivative does not exist due to the absolute value. Thus we can apply standard optimization techniques from Calculus to solve for <math>g_k(x)</math>: find the critical points of <math>g_{k,x}</math> and take the infimum of <math>g_{k,x}</math> evaluated at the critical points. One of these values will always be the original function <math>g</math> evaluated at <math>x</math>, since this corresponds to the critical point <math>y = x</math> for <math>g_{k,x}</math>.
* Let <math>g(x) := x^2</math>.  Then
 
:<math>g_k(x) = \min \left\{ x^2 , \frac{k^2}{2} + k \left| x \pm \frac{k}{2} \right| \right\}.</math>


==Results==
==Results==

Revision as of 00:01, 9 February 2022

(to be filled in)

Motivation

(to be filled in)


Definitions

Let be a metric space. A function is said to be proper if it is not identically equal to , that is, if there exists such that .

For a given function and , its Moreau-Yosida regularization is given by



Examples

  • If , then by definition is constant and .
  • If is not proper, then for all .

Take . If is finite-valued and differentiable, we can explicitly write down . Then for a fixed , the map is continuous everywhere and differentiable everywhere except for when , where the derivative does not exist due to the absolute value. Thus we can apply standard optimization techniques from Calculus to solve for : find the critical points of and take the infimum of evaluated at the critical points. One of these values will always be the original function evaluated at , since this corresponds to the critical point for .

  • Let . Then

Results

Proposition. [1]

  • If is proper and bounded below, so is . Furthermore, is continuous for all .
  • If, in addition, is lower semicontinuous, then for all .
  • In this case, is continuous and bounded and for all .


References

Possible list of references, will fix accordingly

Bauschke-Combette Ch 12.[2]; Santambrogio (6)[3]; Ambrosio-Gigli-Savare (59-61)[4]

  1. Craig, Katy C. Lower Semicontinuity in the Narrow Topology. Math 260J. Univ. of Ca. at Santa Barbara. Winter 2022.
  2. Bauschke, Heinz H. and Patrick L. Combettes. Convex Analysis and Monotone Operator Theory in Hilbert Spaces, 2nd Ed. Ch. 12. Springer, 2017.
  3. Santambrogio, Filippo. Optimal Transport for Applied Mathematicians: Calculus of Variations, PDEs, and Modeling Ch. 1.1. Birkhäuser, 2015.
  4. Ambrosio, Luigi, Nicola Gigli, and Giuseppe Savaré. Gradient Flows in Metric Spaces and in the Space of Probability Measures. Ch. 3.1. Birkhäuser, 2005.