## Optimization problem.

For the discussion of math. Duh.

Moderators: gmalivuk, Moderators General, Prelates

quintopia
Posts: 2906
Joined: Fri Nov 17, 2006 2:53 am UTC
Location: atlanta, ga

### Optimization problem.

What is the supremum of a*x+(1-a)*(1-x)-sqrt(x*x+(1-x)*(1-x))?

For a=.75, I got about -0.161. I can't really see what this has to do with the parameters.

Actually, it would be good enough to know that the answer is always negative.

Please, hints only, as this is for class.
Last edited by quintopia on Sat Apr 25, 2009 1:57 am UTC, edited 1 time in total.

GreedyAlgorithm
Posts: 286
Joined: Tue Aug 22, 2006 10:35 pm UTC
Contact:

### Re: Optimization problem.

Hint #1: b does not appear in the expression you gave.
GENERATION 1-i: The first time you see this, copy it into your sig on any forum. Square it, and then add i to the generation.

skeptical scientist
closed-minded spiritualist
Posts: 6142
Joined: Tue Nov 28, 2006 6:09 am UTC
Location: San Francisco

### Re: Optimization problem.

Hint #2 (actually, this is just a question, but I suspect it may in fact be a hint): do you care about all a and x, or just when they are between 0 and 1?
I'm looking forward to the day when the SNES emulator on my computer works by emulating the elementary particles in an actual, physical box with Nintendo stamped on the side.

"With math, all things are possible." —Rebecca Watson

quintopia
Posts: 2906
Joined: Fri Nov 17, 2006 2:53 am UTC
Location: atlanta, ga

### Re: Optimization problem.

I've reduced the original problem down to one where i only care about [0,1]

NathanielJ
Posts: 882
Joined: Sun Jan 13, 2008 9:04 pm UTC

### Re: Optimization problem.

If all you care about is whether or not it's negative, then just consider it as a function of a and x and find its maximum in the unit square. Standard multivariable calculus extrema-finding tools should work fine, I think.
Homepage: http://www.njohnston.ca
Conway's Game of Life: http://www.conwaylife.com

quintopia
Posts: 2906
Joined: Fri Nov 17, 2006 2:53 am UTC
Location: atlanta, ga

### Re: Optimization problem.

Oh, I'm an idiot , I should have simplified the expression more once I converted it to that form. Turns out it's not even a function in a.

Edit: copied it wrong. yes it is.

quintopia
Posts: 2906
Joined: Fri Nov 17, 2006 2:53 am UTC
Location: atlanta, ga

### Re: Optimization problem.

I got this really nasty expression for the critical point, and plugging it back in isn't really going to tell me much of anything unless I can simplify it a lot. Anyone have an expression simplifier for very long expressions handy?

NathanielJ
Posts: 882
Joined: Sun Jan 13, 2008 9:04 pm UTC

### Re: Optimization problem.

quintopia wrote:I got this really nasty expression for the critical point, and plugging it back in isn't really going to tell me much of anything unless I can simplify it a lot. Anyone have an expression simplifier for very long expressions handy?

Mild spoiler:
Spoiler:
Setting the gradient equal to zero, you get that the derivative of the expression with respect to a is really simple and it gives you that x = 1/2. Thus the only three values of x you care about are x = 0, 1/2, 1. Plug those three values of x into your original expression and you get three expressions in a that are *very* easy to maximize.
Homepage: http://www.njohnston.ca
Conway's Game of Life: http://www.conwaylife.com

quintopia
Posts: 2906
Joined: Fri Nov 17, 2006 2:53 am UTC
Location: atlanta, ga

### Re: Optimization problem.

I was thinking about the original problem in entirely the wrong way. Of course <x,v> is maximized when x is parallel to the unit vector v, and of course <x,v>-||x|| is non-positive.