Optimization With Inequality Constraints: Please Help


by lynxman72
Tags: constraints, inequality, optimization
lynxman72
lynxman72 is offline
#1
Dec9-05, 04:58 PM
P: 16
Hi all, I'm having trouble with the following problem: It was given as a word problem from which to infer the mathematics but basically it is this:

Maximize: f(x,y,z,t,w)=
ln((y^2-x^2)(z^2-t^2)*w^3)+.8x-1.2y-20z/17+14t/17-w^3/(pi^3)

Subject to the constraints:
0<= .5x+y+3z+3y+2.5w<+30
1800<=+130x+150y+200z+70t+110w<=3000

the problem is I don't know how to handle constraints like these, where there is an upper and lower bound, and i'm just confused because it seems like there is a lot going on...the only thing i knew how to do to
1) remark that the max exists because the function is defined on this closed and bounded set 2) compute all of the partials set them equal to zero and find the critical points and hope that I would find the global max and the values at the global max would lie inside the constraints, but this didn't work because when I do this I get the critical point
(2,3,7/3,10/3,pi) which doesn't satisfy the second constraint.
I know that when you have a function defined on a closed bounded set to find the local max you find the crit. points and test the values at the boundaries and just compare but i don't know how to obtain bounds on each individual variable from the constraints or what to do. any help would be appreciated
Phys.Org News Partner Science news on Phys.org
NASA's space station Robonaut finally getting legs
Free the seed: OSSI nurtures growing plants without patent barriers
Going nuts? Turkey looks to pistachios to heat new eco-city
lynxman72
lynxman72 is offline
#2
Dec9-05, 05:02 PM
P: 16
oh sorry by the way I also know that each variable is >=0 from the context of the problem (each variable represents a serving of food). thank you
benorin
benorin is offline
#3
Dec9-05, 09:50 PM
HW Helper
P: 1,007
If you take the first order partials, set them = 0 , and solve the resulting system, you get:

[tex] x = 2, y = 3, z = \frac{10}{3}, t = \frac{7}{3}, w = \pi [/tex]

but these values do not fit the first constraint, namely that

[tex]0\leq \frac{x}{2}+y+3z+3y+\frac{5}{2}w< 30[/tex]

in fact those values of x,y,z,t, and w, will give [itex]\frac{x}{2}+y+3z+3y+\frac{5}{2}w=23+\frac{5}{2}\pi = 30.86 \mbox{ (appox.)}[/itex]

Also note that [itex]f(1,1,6.70000000000000016,1,1)=-\infty[/itex].

lynxman72
lynxman72 is offline
#4
Dec10-05, 12:24 AM
P: 16

Optimization With Inequality Constraints: Please Help


Right, I was just saying all I could figure out how to do was try to find this global max and hope it fit the constraints (it didn't), so any advice on another way to proceed?
benorin
benorin is offline
#5
Dec10-05, 12:56 AM
HW Helper
P: 1,007
Crap, I minimized. Let me look again.

Well, you might try Lagrange Multipliers with the constraints as equalities for the upper and lower bounds of the inequalities.

That version of Lagrange Multipliers goes:

To find the extrema of [itex]f(x,y,z,...)[/itex] subject to the constraints [itex]g_{1}(x,y,z,...)=k_{1}, g_{2}(x,y,z,...)=k_{2}, g_{3}(x,y,z,...)=k_{3},...[/itex] set

[tex]\vec{\nabla} f(x,y,z,...) = \lambda_{1}\vec{\nabla} g_{1}(x,y,z,...) + \lambda_{2}\vec{\nabla} g_{2}(x,y,z,...) + \lambda_{3}\vec{\nabla} g_{3}(x,y,z,...) +\cdot\cdot\cdot [/tex]

and solve as normal.
lynxman72
lynxman72 is offline
#6
Dec10-05, 10:06 AM
P: 16
So what you're saying is then I will have four Lagrange multipliers? I treat it as four constraint equalitites?


Register to reply

Related Discussions
Global optimization subject to constraints Calculus & Beyond Homework 3
Optimization inequality constraint Calculus & Beyond Homework 4
Proof this inequality using Chebyshev's sum inequality Calculus & Beyond Homework 1
Constraints Advanced Physics Homework 5
Constraints Advanced Physics Homework 1