Thanks. What about a method like NonLinearConjugateGradientOptimizer? The
API is not so clear - are there any examples of usage?
If it can be helpful for others, I will collate the information I am
receiving on the general topic of optimizing with commons and post it.
Philippe
On Fri, May 24, 2013 at 10:28 AM, Gilles <gil...@harfang.homelinux.org>wrote:
> On Fri, 24 May 2013 09:52:23 -0400, Mister Mak wrote:
>
>> The multivariate function in my example was indeed flawed. Here is a
>> better example with Thomas' solution:
>>
>> MultivariateFunction g =
>> new MultivariateFunction() {
>> public double value(double[] x) {
>> return Math.exp(-Math.abs(x[0] - 5.4)) *
>> Math.exp(-Math.abs(x[1] - 2.2)) + 1.1;
>> }
>> };
>> SimplexOptimizer optimizerMult = new SimplexOptimizer(1e-3, 1e-6);
>> PointValuePair solutionMult = optimizerMult.optimize(new
>> MaxEval(200), new ObjectiveFunction(g), GoalType.MAXIMIZE, new
>> InitialGuess(new double[]{0, 0}), new MultiDirectionalSimplex(2));
>>
>> System.out.println("Min is: " + solutionMult.getValue() +
>> "\tobtained at:" + Arrays.toString(solutionMult.**getKey()));
>>
>>
>> The above code works. However, introducing constraints like below,
>> doesn't.
>>
>> PointValuePair solutionMult = optimizerMult.optimize(new MaxEval(200), new
>> ObjectiveFunction(g), GoalType.MAXIMIZE, new InitialGuess(new double[]{0,
>> 0}), new MultiDirectionalSimplex(2), new SimpleBounds(new double[]{-10,
>> -15}, new double[]{12, 22}));
>>
>> So which optimizer does accept bounds?
>>
>
> "SimpleBounds" are used by "CMAESOptimizer".
>
> For the other optimizers, bound handling can be mimicked by using the
> MultivariateFunctionMappingAda**pter
> or
> MultivariateFunctionPenaltyAda**pter
> defined in package "o.a.c.m.optim.nonlinear.**scalar".
>
> HTH,
> Gilles
>
>
>
> ------------------------------**------------------------------**---------
> To unsubscribe, e-mail:
> user-unsubscribe@commons.**apache.org<user-unsubscr...@commons.apache.org>
> For additional commands, e-mail: user-h...@commons.apache.org
>
>
