Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: nonlinear optimization problem with constraints
Posted:
Sep 8, 2012 10:43 AM


On 20120908 10:56:42 0300, oercim said:
> On Saturday, September 8, 2012 4:35:17 PM UTC+3, Gordon Sande wrote: >> On 20120908 05:50:07 0300, oercim said: >> >> >> >>> Hello.I have an nonlinear optimization problem with constraints. I >> >>> realy need help. I am using a package problem to solve this problem. >> >>> However, although the program finds a solution without an error( since >> >>> tolarance values are satisfied), when I check the constraints values , >> >>> O see that some constraint values are not satisfactory (some >> >>> constraints are realy close to zero which is satisfactory, but some are >> >>> not close enough to zero). >> >>> There are 5 unknowns and 4 constraints. So this system has infinitly >> >>> many solutions. I just want which solution makes the objevtive function >> >>> minimum. >> >>> I don't have a much optmization background. I just want to determine >> >>> among solutions(where the constraints are close enough to zero) which >> >>> one makes the objective function minimum. As I mentioned, my priority >> >>> is constraints to be satisfied satisfactorily. >> >>> I don't know how to searh this subject. I studied some nonlinear >> >>> optmization resources. However all terminates when the tolareance >> >>> values are close to zero not each of the constraints close to zero. Is >> >>> there a way to control the closenes of the constraint to zero. If there >> >>> is a way, I will be very glad for recomendaton of some resources. At >> >>> least if this subject has an special name in the literature, I want to >> >>> leaarn it.Thanks a lot. Best regards. >> >> >> >> Have you consulted the Guide to Optimization Software at >> >> <http://plato.asu.edu/guide.html>? >> >> >> >> The statements that some constraints are close to zero and others are >> >> not satisfactory does >> >> not sound like you have a clear idea of what constraints are supposed >> >> to do. Are they exact >> >> conditions  always zero? Are they boundaries  so closeness does not >> >> matter? Etc? Is your >> >> objective being used to judge the closeness to the constraints or is it >> >> independent >> >> of the constraints? > My predetermined constraints are exactly zero. However after the > package(matlab's fmincon function) solves the optimization problem, I > checked if the constraints are satisfied with the solution.Then, I see > that > two constraint are satisfied. Their values are "0" , however 2 > constraints have values like 0.01 or 0.1 but not close to zero or zero.
For questions about MatLab functions you are better off asking in a MatLab support group or forum.
An error of 0.1 may be lousy if the plausibe range of the variable is from 1 to +1 but it can be rather good if the plausible range is from 10^15 to +10^15 as it might be if you are looking at national economies measured in Yen. You may know all the details but everyone else is left to trying mind reading. Not a good way to get helpful responses.
Using posting software that tosses in extra blank lines so you response is double spaced is not going to encourage folks to read your postings.
If someone from another course asked you the questions you are trying to pose, would you have a clue what they were talking about? If the answer is NO then work on posing the questions so others have a chance at knowing what is going on.



