On 05/24/2013 12:02 AM, Thomas Neidhart wrote:
> On 05/23/2013 11:17 PM, Mister Mak wrote:
>> Thanks to Thomas and Gilles for their tips. The univariate solution
>> proposed works (for the record:
>> UnivariateFunction f = new UnivariateFunction() {public double
>> value(double x) {return -(x - 5.0) * (x - 5.0) + 2.25; }};
>> BrentOptimizer optimizer = new BrentOptimizer(1e-6, 1e-12);
>> UnivariatePointValuePair solution = optimizer.optimize(new
>> UnivariateObjectiveFunction(f), new MaxEval(100), GoalType.MAXIMIZE, new
>> SearchInterval(-10, 10));
>> System.out.println("Min is: " + solution.getValue() + "\tobtained
>> at: " + solution.getPoint());
>> ).
>>
>> I am now trying to maximize a function of two variables, and can't get the
>> following code running:
>>
>> MultivariateFunction g =
>> new MultivariateFunction() {
>> public double value(double[] x) {
>> return -(x[0] - 5.0) * (x[1] - 5.0) + 2.25;
>> }
>> };
>> SimplexOptimizer optimizerMult = new SimplexOptimizer(1e-6, 1e-12);
>> PointValuePair solutionMult = optimizerMult.optimize(new
>> ObjectiveFunction(g), GoalType.MAXIMIZE, new MultiDirectionalSimplex(100));
>
> You need to set an InitialGuess, like this:
>
> MultivariateFunction g = new MultivariateFunction() {
> public double value(double[] x) {
> return -(x[0] - 5.0) * (x[1] - 5.0) + 2.25;
> }
> };
> 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(solutionMult.getValue());
> System.out.println(Arrays.toString(solutionMult.getKey()));
>
> But this will still not work, as a possible maximum is x[0] = -Infinity
> and x[1] >= 0
>
> I guess in your case you will have to use an optimizer which supports
> bounds to get the maximum for the given function, e.g. the SimplexSolver
> in the linear package.
ok, please ignore my last comment, you do not have a linear problem so
the linear solver will not help you.
Thomas
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
