What is the connection between critical points and global/local optimization?

In summary: Yes, the absolute min value is attained on the "open" set D. Since it is "open", it contains none of its boundary points, so "a" must be a critical point with grad f(a) = 0, am I right?
  • #1
kingwinner
1,270
0
1) http://www.geocities.com/asdfasdf23135/advcal28.JPG

From the assumptions, I think that the mean value theorem and/or the extreme value theorem may be helpful in this problem, but I can't figure out how to apply them to reach the conclusion. Could someone please give me some general hints? Very much appreciated!:smile:
 
Physics news on Phys.org
  • #2
Actually it looks like you want to use Rolle's Theorem (which is a specific case of the mean value theorem anyway). Since f(x)=0 for all x then between any x1 and x2 you have a point c such that f'(c)=0.
 
  • #3
No, "Rolle's theorem" applies to functions on R, not Rn. You can use the "Extreme Value theorem" and "mimic" the proof or Rolle's theorem. There are 3 possibliities:
1) There are positive values of f(x,y) inside the boundary
2) There are negative values of f(x,y) inside the boundary
3) There are neither positive nor negative values of f(x,y) inside the boundary

What does the extreme value theorem tell you in each of those cases?
 
  • #4
HallsofIvy said:
No, "Rolle's theorem" applies to functions on R, not Rn. You can use the "Extreme Value theorem" and "mimic" the proof or Rolle's theorem. There are 3 possibliities:
1) There are positive values of f(x,y) inside the boundary
2) There are negative values of f(x,y) inside the boundary
3) There are neither positive nor negative values of f(x,y) inside the boundary

What does the extreme value theorem tell you in each of those cases?
Thanks! Your hints are helpful!

By extreme value theorem (EVT), since f is continuous and the closure of D is compact, there exists an absolute max value and an absolute min value on the closure of D.

If f(x) is identically 0 on closure of D, then any a on D will do.
If f(x) is not identically 0 on closure of D, since f(x)=0 on boundary of D, we must have either f(x)>0 for some x on D or f(x)<0 for some x on D
Let's consider the case f(x)<0 for some x on D. Absolute min value must be <0 and since f(x)=0 on boundary of D, this min value must occur on the open set D. Say f(a), a E D is the absolute min value.

Now, does this imply that f(a) is a local min and that grad f(a)=0? Why or why not?
 
  • #5
Continuing with the last post:

f(a), the absolute min, is attained on the "open" set D. Since it is "open", it contains none of its boundary points, so "a" must be a critical point with grad f(a) = 0, am I right?
 

FAQ: What is the connection between critical points and global/local optimization?

What is local optimization and how does it differ from global optimization?

Local optimization is a method used to find the best solution to a problem within a specific range of values. It is limited to a small area of the problem space. On the other hand, global optimization aims to find the best solution for the entire problem space, without being restricted to a specific range of values.

What are some common techniques used in local optimization?

Some common techniques used in local optimization include gradient descent, hill climbing, and simulated annealing. These methods involve iteratively adjusting parameters to improve the solution until an optimal value is reached.

How does the choice of objective function impact the effectiveness of local optimization?

The objective function, which is the measure of how well the solution satisfies the problem's requirements, has a significant impact on the effectiveness of local optimization. A well-defined objective function that accurately represents the problem can lead to a more efficient and accurate solution.

Can local optimization be used for non-convex problems?

Yes, local optimization can be used for non-convex problems. However, it may not always guarantee finding the global optimum, as it may get stuck in a local optimum. In such cases, global optimization techniques may be more suitable.

What are some applications of local and global optimization?

Local and global optimization have various applications in fields such as engineering, economics, finance, and machine learning. They are used to optimize parameters in complex systems, improve resource allocation, and train machine learning models, among others.

Similar threads

Back
Top