# Derivation of Lagrange Multipliers Method

1. Jun 5, 2010

Hey folks. I have some more or less qualitative questions regarding optimization problems via Lagrange multipliers. I am following the http://en.wikipedia.org/wiki/Lagrange_multipliers" [Broken] on this one and I am just a little confused by their wording.

In the first section titled "Introduction," they have the following:

1) My first question (which might seem silly) is: What is g(x,y) ? Is it a line? Or a surface? It looks like a line in their picture but I am used to functions of the forms g(x,y) being representative of surfaces; though I feel like it could be ambiguous.

2) Second, when they say: "Suppose we walk along the contour line with g = c." They don't mean that g = c IS the contour line right? They are saying that AT g(x,y) one can follow a contour line. It just seems like if g(x,y) = c is not a surface, then the whole idea of a contour is a little weird to me. So perhaps if someone can answer my first question this would make more sense to me.

Sorry if these seem stupid, but it has been awhile.

Last edited by a moderator: May 4, 2017
2. Jun 5, 2010

### LCKurtz

3. Jun 5, 2010

Hey LCKurtz

I am sketching this right now. Sketching level sets was never really my strong suit.
Is it as simple as letting c take on a certain value and then plotting that line? For example, let c = 1/2 and then plot the line y = 1/2 - 2x ? I am going to assume that the answer is yes and proceed.

And for this example, I assume that we are maximizing 'T' subject to the constraint that the point falls on the wire.

EDIT: Okay. I have plotted the circle and the line y = c - 2x for various values of c. I can see that 'T' would be maximized (and minimized) when the line y = c - 2x is tangent to the circle. I am having a couple of issues with this now.

1) Not sure how to generalize this idea beyond the simple circle constraint. That is if both my maximizing function f and my constraining function g are arbitrarily shaped surfaces, will the condition that g is 'tangent to' f be sufficient to say that f is at a maximum there?

2) I am not sure how to word this without sounding like an idiot but here goes anyway: The function T = y + 2x = c that you gave, is that a surface? That is, isn't it T(x,y) = y + 2x - c ? Or is it just T(x,y) = y + 2x ? I don't know why this is so confusing right now

Last edited: Jun 5, 2010
4. Jun 5, 2010

### vela

Staff Emeritus
It could be a minimum as well.
T(x,y)=y+2x describes a surface. Think of T as the z coordinate. For each point in the xy plane, there's an associated temperature which you can think of as the height of the surface above the point (x,y). When you choose a specific value of T, you're looking at the intersection of that surface with the plane T=c.

5. Jun 6, 2010

### LCKurtz

Yes

This is a 2D example.

T = y+2x is the temperature at a point (x,y) in the plane. I don't find it helpful to interpret it as a surface although it can be thought of that way. The level curves in this example are straight lines in the plane on which the temperature is constant.: y+2x = c. It wouldn't matter if the level curves had other shapes. Try drawing a family of hyperbolas. As long as everything is smooth the level curves will be tangent at the extremes.

6. Jun 8, 2010

For anyone who is curious, I plotted T(x,y) = y + 2x and g(x,y) = x^2 + y^2 along with the plane z(x,y) = 1. The intersection of z(x,y,) = 1 with g(x,y) = x^2 + y^2 gives the level set g(x,y) = x^2 + y^2 = 1. Here are a couple of views:

and another view

And and one more

7. Jun 9, 2010

### LCKurtz

Pretty pictures, but I don't think they add to the understanding of what is going on. The example problem is given temperature T(x,y) = x+2y in the plane, find the hottest point on the wire represented by the unit circle x2+y2 = 1. The following plot shows the level curves for T varying from -3 to 3 including, in red, T =sqrt(5) which obviously gives the max.

Maple doesn't label the contours, but in this example the y intercepts of the contour lines give T for the lines.

8. Jun 9, 2010