Re: fmincon search direction

Do u use the same algorithm

Nasim wrote:
>
>
> sachin porandla wrote:
>>
>>
>> I dont have analytical functions for either the objective or
the
>> constraint functions. I only get those values from a simulation
>> program.But they are continuous and differentiable. I am not
sure
>> how
>> noisy there are.
>>
>> I recently downloaded Tomlab. So could you tell what solvers
fit
>> for
>> the constrained non linear problems
>>
>> sachin
>>
>> Marcelo Marazzi wrote:
>>>
>>>
>>>
>>> sachin porandla wrote:
>>>> I try to maximize a function. This is a medium-scale
> problem.
>> It
>>> gets
>>>> maximized but to a little extent because only 2
varaibles
>> change
>>> out
>>>> of 6 while the other remain unchanged.
>>>>
>>>> I get this message "Optimization terminated: no
feasible
>> solution
>>>> found. Magnitude of search direction less than
> 2*options.TolX
>> but
>>>> constraints are not satisfied." with an EXITFLAG=-2.I
> guess
>> this
>>>> means an infeasible initial point.So i changed the
> initial
>> points
>>> but
>>>> still only few variables change.
>>> This means that a feasible point was not found. It could be
>> either
>>> that there are feasible points but the solver couldn't find
> any,
>> or
>>> that the constraints don't admit any feasible solution. Do
> you
>> know
>>> a feasible point? If you do, try that one as x0.
>>>
>>> What kind of problem is this? Are the objective and the
>>> constraints continuous and differentiable (that is,
smooth)?
>>> Are your functions noisy? Depending on the answer to these
>>> questions, fmincon may or may not be the right solver.
>>> (E.g., if your functions are nonsmooth, a solver in the
>>> Genetic Algorithm and Pattern Search Tbx may be more
>>> appropriate. Or, if your functions are noisy, increasing
>>> DiffMinChange may help).
>>>
>>> Feel free to send your files to tech support,
>>> <http://www.mathworks.com/support/contact_us/>
>>>
>>> -marcelo
>>>
>>>> I tried using fminimax for this problem.That behaved
more
>> worse
>>> than
>>>> fmincon. Even those two variables changed by a little
> amount.
>> I
>>> use
>>>> the default options for both fmincon and fminimax.
>>>>
>>>> What should I do??
>>>>
>>>> Could you suggest any other algorithms in Matlab Opt.
> toolbox
>> or
>>>> Tomlab for solving constrained non linear problems.
>>>>
>>>> sachin
>>>>
>>>> Marcelo Marazzi wrote:
>>>>
>>>>>
>>>>>What is the result of the overall optimization?
>>>>>Is the solver finding a minimizer? In other words,
>>>>>what is the final message and exitflag value?
>>>>>
>>>>>Do you observe the same behavior starting from
>>>>>totally different starting points?
>>>>>
>>>>>-marcelo
>>>>>
>>>>>sachin porandla wrote:
>>>>>
>>>>>>Hi..
>>>>>>I am using fmincon for a maximization problem.
> This
> is
>> a
>> six
>>>>>
>>>>>variable
>>>>>
>>>>>>problem.
>>>>>>
>>>>>>My problem is that only three variables are
> changing
>> i.e. SD
>>>>>
>>>>>(search
>>>>>
>>>>>>direction) variable is 0 for three other
> variables.
> I
>> dont know
>>>>>
>>>>>the
>>>>>
>>>>>>reason.
>>>>>>
>>>>>>Could any one guess what is the problem. Is
there
> any
>> way to
>>>>>
>>>>>correct
>>>>>
>>>>>>this? Are there any other programs avaliable for
>> solving
>>>>>
>>>>>constrained
>>>>>
>>>>>>non-linear problems online.
>>>>>>
>>>>>>Thanks
>>>>>>sachin
>>>>>
>>>
>
> I have the same problem!
.

Relevant Pages

  • Re: fmincon search direction
    ... >> sachin porandla wrote: ... >>> still only few variables change. ... >> that the constraints don't admit any feasible solution. ... fmincon may or may not be the right solver. ...
    (comp.soft-sys.matlab)
  • Re: A Two-Level SOLVER ??
    ... I'm running Excel Solver from a macro. ... ....Subject to constraints: B18=1 ... The correct formulation of the problem should be, I think, something like: ... cells or a vector of cells. ...
    (microsoft.public.excel.programming)
  • Re: fmincon search direction
    ... Magnitude of search direction less than 2*options.TolX but constraints are not satisfied." ... fmincon may or may not be the right solver. ...
    (comp.soft-sys.matlab)
  • Re: projection of a point to {x | b1 <= Ax <= b2}?
    ... But instead of using QP I intended to use constrained LLSP solver, ... BTW I intend to involve it for to speedup my NLP/NSP sover ralg ... Suppose I have set of linear constraints ... If you want the sortest L_1 distance that is linear programming. ...
    (sci.math.num-analysis)
  • Re: Coding complex VHDL testbenches
    ... variable RV: RandomPType; ... The limitation to this case is the distribution of values is not ... that the solver gives a uniform distribution across possible values. ... You can specify some constraints in the class, ...
    (comp.lang.vhdl)
Loading