Math 2.0

Calculus of variations and optimal control; optimization Definition

Weak and strong extremum

Builds on Admissible functionFunctionalNorm

Definition

Let a<ba < b, let π’œβŠ†C1([a,b])\mathcal{A} \subseteq C^1([a, b]) be a class of admissible functions, and let J ⁣:π’œβ†’β„J \colon \mathcal{A} \to \R be a functional. For y∈C1([a,b])y \in C^1([a, b]) use the norms βˆ₯yβˆ₯0=max⁑[a,b]∣y∣,βˆ₯yβˆ₯1=max⁑[a,b]∣y∣+max⁑[a,b]∣yβ€²βˆ£.\norm{y}_0 = \max_{[a, b]} \abs{y}, \qquad \norm{y}_1 = \max_{[a, b]} \abs{y} + \max_{[a, b]} \abs{y'}. Let y0βˆˆπ’œy_0 \in \mathcal{A}.

  1. y0y_0 is a strong local minimum of JJ if there is Ξ΅>0\varepsilon > 0 with J(y0)≀J(y)J(y_0) \le J(y) for all yβˆˆπ’œy \in \mathcal{A} with βˆ₯yβˆ’y0βˆ₯0<Ξ΅\norm{y - y_0}_0 < \varepsilon.
  2. y0y_0 is a weak local minimum of JJ if there is Ξ΅>0\varepsilon > 0 with J(y0)≀J(y)J(y_0) \le J(y) for all yβˆˆπ’œy \in \mathcal{A} with βˆ₯yβˆ’y0βˆ₯1<Ξ΅\norm{y - y_0}_1 < \varepsilon.

Strong and weak local maxima are defined with J(y0)β‰₯J(y)J(y_0) \ge J(y). A strong (weak) local extremum is a strong (weak) local minimum or maximum.

Remarks

Since βˆ₯yβˆ₯0≀βˆ₯yβˆ₯1\norm{y}_0 \le \norm{y}_1, every strong local minimum is a weak local minimum.

Example: let J(y)=∫01yβ€²2 dxJ(y) = \int_0^1 y'^2 \, dx on π’œ={y∈C1([0,1]):y(0)=0,Β y(1)=1}\mathcal{A} = \set{y \in C^1([0, 1]) : y(0) = 0, \ y(1) = 1}, and y0(x)=xy_0(x) = x. For yβˆˆπ’œy \in \mathcal{A} write yβ€²=1+Ξ·y' = 1 + \eta; then ∫01η dx=0\int_0^1 \eta \, dx = 0, so J(y)=1+∫01Ξ·2 dxβ‰₯J(y0)J(y) = 1 + \int_0^1 \eta^2 \, dx \ge J(y_0). So y0y_0 is a global, hence strong, minimum.

Non-example: let J(y)=∫01yβ€²3 dxJ(y) = \int_0^1 y'^3 \, dx on the same class, and y0(x)=xy_0(x) = x. With yβ€²=1+Ξ·y' = 1 + \eta one gets J(y)=1+∫01Ξ·2(3+Ξ·) dxJ(y) = 1 + \int_0^1 \eta^2 (3 + \eta) \, dx, which is at least 11 when ∣η∣<3\abs{\eta} < 3; so y0y_0 is a weak local minimum. It is not a strong local minimum. Fix h>0h > 0, let NN be large and Ξ΄=h/N\delta = h/N, and let yy have slope βˆ’N-N on [0,Ξ΄][0, \delta] and slope (1+h)/(1βˆ’Ξ΄)(1 + h)/(1 - \delta) on [Ξ΄,1][\delta, 1], with corners rounded off. Then βˆ₯yβˆ’y0βˆ₯0≀h+Ξ΄\norm{y - y_0}_0 \le h + \delta, while J(y)J(y) is close to βˆ’N2h+(1+h)3/(1βˆ’Ξ΄)2-N^2 h + (1 + h)^3/(1 - \delta)^2, which tends to βˆ’βˆž-\infty as Nβ†’βˆžN \to \infty.

Reading path

To understand this page from the ground up, read these 4 pages in order. The foundations come first.

Show the path as a graph
Each page uses the pages above it. The colour of the bar on a box shows how strong the evidence is, and the box also gives it as text. The list has the same pages.
  • no evidence level
  1. Admissible function concept
  2. Functional concept
  3. Norm concept
  4. Weak and strong extremum definition

Further reading

  1. Web https://en.wikipedia.org/wiki/Calculus_of_variations
About this page
Tag4R6XZSDH3E
URIm2://4R6XZSDH3E
Kinddefinition
WritingW0 stub
MSC49K05, 49K99
DefinesWeak and strong extremum
UsesAdmissible function, Functional, Norm

Markdown Β· JSON Β· Edit on GitHub

What this page needs for W1

  • an inventory row: the row of the inventory that the node accounts for (0 of 1)
  • a source with a locator (0 of 1)

The writing levels say how complete the text of a page is. They do not measure its truth.

What links here

No page links to this page yet.