Java Code Examples for org.apache.commons.math3.random.JDKRandomGenerator#setSeed()

@Test
public void testRosenbrock() {
    Rosenbrock rosenbrock = new Rosenbrock();
    SimplexOptimizer underlying
        = new SimplexOptimizer(new SimpleValueChecker(-1, 1.0e-3));
    NelderMeadSimplex simplex = new NelderMeadSimplex(new double[][] {
            { -1.2,  1.0 }, { 0.9, 1.2 } , {  3.5, -2.3 }
        });
    underlying.setSimplex(simplex);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(16069223052l);
    RandomVectorGenerator generator =
        new UncorrelatedRandomVectorGenerator(2, new GaussianRandomGenerator(g));
    MultivariateMultiStartOptimizer optimizer =
        new MultivariateMultiStartOptimizer(underlying, 10, generator);
    PointValuePair optimum =
        optimizer.optimize(1100, rosenbrock, GoalType.MINIMIZE, new double[] { -1.2, 1.0 });

    Assert.assertEquals(rosenbrock.getCount(), optimizer.getEvaluations());
    Assert.assertTrue(optimizer.getEvaluations() > 900);
    Assert.assertTrue(optimizer.getEvaluations() < 1200);
    Assert.assertTrue(optimum.getValue() < 8.0e-4);
}
 

@Test
public void testSinMin() {
    UnivariateFunction f = new Sin();
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 10, g);
    optimizer.optimize(300, f, GoalType.MINIMIZE, -100.0, 100.0);
    UnivariatePointValuePair[] optima = optimizer.getOptima();
    for (int i = 1; i < optima.length; ++i) {
        double d = (optima[i].getPoint() - optima[i-1].getPoint()) / (2 * FastMath.PI);
        Assert.assertTrue(FastMath.abs(d - FastMath.rint(d)) < 1.0e-8);
        Assert.assertEquals(-1.0, f.value(optima[i].getPoint()), 1.0e-10);
        Assert.assertEquals(f.value(optima[i].getPoint()), optima[i].getValue(), 1.0e-10);
    }
    Assert.assertTrue(optimizer.getEvaluations() > 200);
    Assert.assertTrue(optimizer.getEvaluations() < 300);
}
 

@Test
public void testBadFunction() {
    UnivariateFunction f = new UnivariateFunction() {
            public double value(double x) {
                if (x < 0) {
                    throw new LocalException();
                }
                return 0;
            }
        };
    UnivariateOptimizer underlying = new BrentOptimizer(1e-9, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(4312000053L);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 5, g);
 
    try {
        optimizer.optimize(300, f, GoalType.MINIMIZE, -0.3, -0.2);
        Assert.fail();
    } catch (LocalException e) {
        // Expected.
    }

    // Ensure that the exception was thrown because no optimum was found.
    Assert.assertTrue(optimizer.getOptima()[0] == null);
}
 

@Test
public void testBadFunction() {
    UnivariateFunction f = new UnivariateFunction() {
            public double value(double x) {
                if (x < 0) {
                    throw new LocalException();
                }
                return 0;
            }
        };
    UnivariateOptimizer underlying = new BrentOptimizer(1e-9, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(4312000053L);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 5, g);
 
    try {
        optimizer.optimize(300, f, GoalType.MINIMIZE, -0.3, -0.2);
        Assert.fail();
    } catch (LocalException e) {
        // Expected.
    }

    // Ensure that the exception was thrown because no optimum was found.
    Assert.assertTrue(optimizer.getOptima()[0] == null);
}
 

@Test
public void testBadFunction() {
    UnivariateFunction f = new UnivariateFunction() {
            public double value(double x) {
                if (x < 0) {
                    throw new LocalException();
                }
                return 0;
            }
        };
    UnivariateOptimizer underlying = new BrentOptimizer(1e-9, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(4312000053L);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 5, g);
 
    try {
        optimizer.optimize(300, f, GoalType.MINIMIZE, -0.3, -0.2);
        Assert.fail();
    } catch (LocalException e) {
        // Expected.
    }

    // Ensure that the exception was thrown because no optimum was found.
    Assert.assertTrue(optimizer.getOptima()[0] == null);
}
 

@Test
public void testSinMin() {
    UnivariateFunction f = new SinFunction();
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 10, g);
    optimizer.optimize(300, f, GoalType.MINIMIZE, -100.0, 100.0);
    UnivariatePointValuePair[] optima = optimizer.getOptima();
    for (int i = 1; i < optima.length; ++i) {
        double d = (optima[i].getPoint() - optima[i-1].getPoint()) / (2 * FastMath.PI);
        Assert.assertTrue(FastMath.abs(d - FastMath.rint(d)) < 1.0e-8);
        Assert.assertEquals(-1.0, f.value(optima[i].getPoint()), 1.0e-10);
        Assert.assertEquals(f.value(optima[i].getPoint()), optima[i].getValue(), 1.0e-10);
    }
    Assert.assertTrue(optimizer.getEvaluations() > 200);
    Assert.assertTrue(optimizer.getEvaluations() < 300);
}
 

@Test
public void testQuinticMin() {
    // The quintic function has zeros at 0, +-0.5 and +-1.
    // The function has extrema (first derivative is zero) at 0.27195613 and 0.82221643,
    UnivariateFunction f = new QuinticFunction();
    UnivariateOptimizer underlying = new BrentOptimizer(1e-9, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(4312000053L);
    MultiStartUnivariateOptimizer optimizer = new MultiStartUnivariateOptimizer(underlying, 5, g);

    UnivariatePointValuePair optimum
        = optimizer.optimize(new MaxEval(300),
                             new UnivariateObjectiveFunction(f),
                             GoalType.MINIMIZE,
                             new SearchInterval(-0.3, -0.2));
    Assert.assertEquals(-0.27195613, optimum.getPoint(), 1e-9);
    Assert.assertEquals(-0.0443342695, optimum.getValue(), 1e-9);

    UnivariatePointValuePair[] optima = optimizer.getOptima();
    for (int i = 0; i < optima.length; ++i) {
        Assert.assertEquals(f.value(optima[i].getPoint()), optima[i].getValue(), 1e-9);
    }
    Assert.assertTrue(optimizer.getEvaluations() >= 50);
    Assert.assertTrue(optimizer.getEvaluations() <= 100);
}
 

@Test
public void testBadFunction() {
    UnivariateFunction f = new UnivariateFunction() {
            public double value(double x) {
                if (x < 0) {
                    throw new LocalException();
                }
                return 0;
            }
        };
    UnivariateOptimizer underlying = new BrentOptimizer(1e-9, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(4312000053L);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 5, g);
 
    try {
        optimizer.optimize(300, f, GoalType.MINIMIZE, -0.3, -0.2);
        Assert.fail();
    } catch (LocalException e) {
        // Expected.
    }

    // Ensure that the exception was thrown because no optimum was found.
    Assert.assertTrue(optimizer.getOptima()[0] == null);
}
 

@Test
public void testSinMin() {
    UnivariateFunction f = new Sin();
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    UnivariateMultiStartOptimizer<UnivariateFunction> optimizer =
        new UnivariateMultiStartOptimizer<UnivariateFunction>(underlying, 10, g);
    optimizer.optimize(300, f, GoalType.MINIMIZE, -100.0, 100.0);
    UnivariatePointValuePair[] optima = optimizer.getOptima();
    for (int i = 1; i < optima.length; ++i) {
        double d = (optima[i].getPoint() - optima[i-1].getPoint()) / (2 * FastMath.PI);
        Assert.assertTrue(FastMath.abs(d - FastMath.rint(d)) < 1.0e-8);
        Assert.assertEquals(-1.0, f.value(optima[i].getPoint()), 1.0e-10);
        Assert.assertEquals(f.value(optima[i].getPoint()), optima[i].getValue(), 1.0e-10);
    }
    Assert.assertTrue(optimizer.getEvaluations() > 200);
    Assert.assertTrue(optimizer.getEvaluations() < 300);
}
 

/**
 * Test demonstrating that the user exception is finally thrown if none
 * of the runs succeed.
 */
@Test(expected=TestException.class)
public void testNoOptimum() {
    JacobianMultivariateVectorOptimizer underlyingOptimizer
        = new GaussNewtonOptimizer(true, new SimpleVectorValueChecker(1e-6, 1e-6));
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(12373523445l);
    RandomVectorGenerator generator
        = new UncorrelatedRandomVectorGenerator(1, new GaussianRandomGenerator(g));
    MultiStartMultivariateVectorOptimizer optimizer
        = new MultiStartMultivariateVectorOptimizer(underlyingOptimizer, 10, generator);
    optimizer.optimize(new MaxEval(100),
                       new Target(new double[] { 0 }),
                       new Weight(new double[] { 1 }),
                       new InitialGuess(new double[] { 0 }),
                       new ModelFunction(new MultivariateVectorFunction() {
                               public double[] value(double[] point) {
                                   throw new TestException();
                               }
                           }));
}
 

@Test
public void testRosenbrock() {
    Rosenbrock rosenbrock = new Rosenbrock();
    SimplexOptimizer underlying
        = new SimplexOptimizer(new SimpleValueChecker(-1, 1e-3));
    NelderMeadSimplex simplex = new NelderMeadSimplex(new double[][] {
            { -1.2,  1.0 },
            { 0.9, 1.2 } ,
            {  3.5, -2.3 }
        });
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(16069223052l);
    RandomVectorGenerator generator
        = new UncorrelatedRandomVectorGenerator(2, new GaussianRandomGenerator(g));
    int nbStarts = 10;
    MultiStartMultivariateOptimizer optimizer
        = new MultiStartMultivariateOptimizer(underlying, nbStarts, generator);
    PointValuePair optimum
        = optimizer.optimize(new MaxEval(1100),
                             new ObjectiveFunction(rosenbrock),
                             GoalType.MINIMIZE,
                             simplex,
                             new InitialGuess(new double[] { -1.2, 1.0 }));
    Assert.assertEquals(nbStarts, optimizer.getOptima().length);

    Assert.assertEquals(rosenbrock.getCount(), optimizer.getEvaluations());
    Assert.assertTrue(optimizer.getEvaluations() > 900);
    Assert.assertTrue(optimizer.getEvaluations() < 1200);
    Assert.assertTrue(optimum.getValue() < 5e-5);
}
 

@Test(expected=MathIllegalStateException.class)
public void testMissingSearchInterval() {
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    MultiStartUnivariateOptimizer optimizer = new MultiStartUnivariateOptimizer(underlying, 10, g);
    optimizer.optimize(new MaxEval(300),
                       new UnivariateObjectiveFunction(new Sin()),
                       GoalType.MINIMIZE);
}
 

@Test(expected=MathIllegalStateException.class)
public void testMissingMaxEval() {
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    MultiStartUnivariateOptimizer optimizer = new MultiStartUnivariateOptimizer(underlying, 10, g);
    optimizer.optimize(new UnivariateObjectiveFunction(new Sin()),
                       GoalType.MINIMIZE,
                       new SearchInterval(-1, 1));
}
 

@Test
public void testRosenbrock() {
    Rosenbrock rosenbrock = new Rosenbrock();
    SimplexOptimizer underlying
        = new SimplexOptimizer(new SimpleValueChecker(-1, 1e-3));
    NelderMeadSimplex simplex = new NelderMeadSimplex(new double[][] {
            { -1.2,  1.0 },
            { 0.9, 1.2 } ,
            {  3.5, -2.3 }
        });
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(16069223052l);
    RandomVectorGenerator generator
        = new UncorrelatedRandomVectorGenerator(2, new GaussianRandomGenerator(g));
    MultiStartMultivariateOptimizer optimizer
        = new MultiStartMultivariateOptimizer(underlying, 10, generator);
    PointValuePair optimum
        = optimizer.optimize(new MaxEval(1100),
                             new ObjectiveFunction(rosenbrock),
                             GoalType.MINIMIZE,
                             simplex,
                             new InitialGuess(new double[] { -1.2, 1.0 }));

    Assert.assertEquals(rosenbrock.getCount(), optimizer.getEvaluations());
    Assert.assertTrue(optimizer.getEvaluations() > 900);
    Assert.assertTrue(optimizer.getEvaluations() < 1200);
    Assert.assertTrue(optimum.getValue() < 8e-4);
}
 

@Test(expected=MathIllegalStateException.class)
public void testMissingMaxEval() {
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    MultiStartUnivariateOptimizer optimizer = new MultiStartUnivariateOptimizer(underlying, 10, g);
    optimizer.optimize(new UnivariateObjectiveFunction(new Sin()),
                       GoalType.MINIMIZE,
                       new SearchInterval(-1, 1));
}
 

@Test
public void testTrivial() {
    LinearProblem problem =
        new LinearProblem(new double[][] { { 2 } }, new double[] { 3 });
    DifferentiableMultivariateVectorOptimizer underlyingOptimizer =
        new GaussNewtonOptimizer(true,
                                 new SimpleVectorValueChecker(1.0e-6, 1.0e-6));
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(16069223052l);
    RandomVectorGenerator generator =
        new UncorrelatedRandomVectorGenerator(1, new GaussianRandomGenerator(g));
    DifferentiableMultivariateVectorMultiStartOptimizer optimizer =
        new DifferentiableMultivariateVectorMultiStartOptimizer(underlyingOptimizer,
                                                                   10, generator);

    // no optima before first optimization attempt
    try {
        optimizer.getOptima();
        Assert.fail("an exception should have been thrown");
    } catch (MathIllegalStateException ise) {
        // expected
    }
    PointVectorValuePair optimum =
        optimizer.optimize(100, problem, problem.target, new double[] { 1 }, new double[] { 0 });
    Assert.assertEquals(1.5, optimum.getPoint()[0], 1.0e-10);
    Assert.assertEquals(3.0, optimum.getValue()[0], 1.0e-10);
    PointVectorValuePair[] optima = optimizer.getOptima();
    Assert.assertEquals(10, optima.length);
    for (int i = 0; i < optima.length; ++i) {
        Assert.assertEquals(1.5, optima[i].getPoint()[0], 1.0e-10);
        Assert.assertEquals(3.0, optima[i].getValue()[0], 1.0e-10);
    }
    Assert.assertTrue(optimizer.getEvaluations() > 20);
    Assert.assertTrue(optimizer.getEvaluations() < 50);
    Assert.assertEquals(100, optimizer.getMaxEvaluations());
}
 

@Test(expected=MathIllegalStateException.class)
public void testMissingMaxEval() {
    UnivariateOptimizer underlying = new BrentOptimizer(1e-10, 1e-14);
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(44428400075l);
    MultiStartUnivariateOptimizer optimizer = new MultiStartUnivariateOptimizer(underlying, 10, g);
    optimizer.optimize(new UnivariateObjectiveFunction(new Sin()),
                       GoalType.MINIMIZE,
                       new SearchInterval(-1, 1));
}
 

@Test
public void testIssue914() {
    LinearProblem problem = new LinearProblem(new double[][] { { 2 } }, new double[] { 3 });
    JacobianMultivariateVectorOptimizer underlyingOptimizer =
            new GaussNewtonOptimizer(true, new SimpleVectorValueChecker(1e-6, 1e-6)) {
        @Override
        public PointVectorValuePair optimize(OptimizationData... optData) {
            // filter out simple bounds, as they are not supported
            // by the underlying optimizer, and we don't really care for this test
            OptimizationData[] filtered = optData.clone();
            for (int i = 0; i < filtered.length; ++i) {
                if (filtered[i] instanceof SimpleBounds) {
                    filtered[i] = null;
                }
            }
            return super.optimize(filtered);
        }
    };
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(16069223052l);
    RandomVectorGenerator generator =
            new UncorrelatedRandomVectorGenerator(1, new GaussianRandomGenerator(g));
    MultiStartMultivariateVectorOptimizer optimizer =
            new MultiStartMultivariateVectorOptimizer(underlyingOptimizer, 10, generator);

    optimizer.optimize(new MaxEval(100),
                       problem.getModelFunction(),
                       problem.getModelFunctionJacobian(),
                       problem.getTarget(),
                       new Weight(new double[] { 1 }),
                       new InitialGuess(new double[] { 0 }),
                       new SimpleBounds(new double[] { -1.0e-10 }, new double[] {  1.0e-10 }));
    PointVectorValuePair[] optima = optimizer.getOptima();
    // only the first start should have succeeded
    Assert.assertEquals(1, optima.length);

}
 

@Test
public void testCircleFitting() {
    CircleScalar circle = new CircleScalar();
    circle.addPoint( 30.0,  68.0);
    circle.addPoint( 50.0,  -6.0);
    circle.addPoint(110.0, -20.0);
    circle.addPoint( 35.0,  15.0);
    circle.addPoint( 45.0,  97.0);
    // TODO: the wrapper around NonLinearConjugateGradientOptimizer is a temporary hack for
    // version 3.1 of the library. It should be removed when NonLinearConjugateGradientOptimizer
    // will officially be declared as implementing MultivariateDifferentiableOptimizer
    GradientMultivariateOptimizer underlying
        = new NonLinearConjugateGradientOptimizer(NonLinearConjugateGradientOptimizer.Formula.POLAK_RIBIERE,
                                                  new SimpleValueChecker(1e-10, 1e-10));
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(753289573253l);
    RandomVectorGenerator generator
        = new UncorrelatedRandomVectorGenerator(new double[] { 50, 50 },
                                                new double[] { 10, 10 },
                                                new GaussianRandomGenerator(g));
    int nbStarts = 10;
    MultiStartMultivariateOptimizer optimizer
        = new MultiStartMultivariateOptimizer(underlying, nbStarts, generator);
    PointValuePair optimum
        = optimizer.optimize(new MaxEval(1000),
                             circle.getObjectiveFunction(),
                             circle.getObjectiveFunctionGradient(),
                             GoalType.MINIMIZE,
                             new InitialGuess(new double[] { 98.680, 47.345 }));
    Assert.assertEquals(1000, optimizer.getMaxEvaluations());
    PointValuePair[] optima = optimizer.getOptima();
    Assert.assertEquals(nbStarts, optima.length);
    for (PointValuePair o : optima) {
        // we check the results of all intermediate restarts here (there are 10 such results)
        Vector2D center = new Vector2D(o.getPointRef()[0], o.getPointRef()[1]);
        Assert.assertTrue(69.9592 < circle.getRadius(center));
        Assert.assertTrue(69.9602 > circle.getRadius(center));
        Assert.assertTrue(96.0745 < center.getX());
        Assert.assertTrue(96.0762 > center.getX());
        Assert.assertTrue(48.1344 < center.getY());
        Assert.assertTrue(48.1354 > center.getY());
    }

    Assert.assertTrue(optimizer.getEvaluations() > 850);
    Assert.assertTrue(optimizer.getEvaluations() < 900);

    Assert.assertEquals(3.1267527, optimum.getValue(), 1e-8);
}
 

@Test
public void testCircleFitting() {
    CircleScalar circle = new CircleScalar();
    circle.addPoint( 30.0,  68.0);
    circle.addPoint( 50.0,  -6.0);
    circle.addPoint(110.0, -20.0);
    circle.addPoint( 35.0,  15.0);
    circle.addPoint( 45.0,  97.0);
    // TODO: the wrapper around NonLinearConjugateGradientOptimizer is a temporary hack for
    // version 3.1 of the library. It should be removed when NonLinearConjugateGradientOptimizer
    // will officially be declared as implementing MultivariateDifferentiableOptimizer
    MultivariateDifferentiableOptimizer underlying =
            new MultivariateDifferentiableOptimizer() {

        private final NonLinearConjugateGradientOptimizer cg =
                new NonLinearConjugateGradientOptimizer(ConjugateGradientFormula.POLAK_RIBIERE,
                                                        new SimpleValueChecker(1.0e-10, 1.0e-10));
        public PointValuePair optimize(int maxEval,
                                       MultivariateDifferentiableFunction f,
                                       GoalType goalType,
                                       double[] startPoint) {
            return cg.optimize(maxEval, f, goalType, startPoint);
        }

        public int getMaxEvaluations() {
            return cg.getMaxEvaluations();
        }

        public int getEvaluations() {
            return cg.getEvaluations();
        }

        public ConvergenceChecker<PointValuePair> getConvergenceChecker() {
            return cg.getConvergenceChecker();
        }
    };
    JDKRandomGenerator g = new JDKRandomGenerator();
    g.setSeed(753289573253l);
    RandomVectorGenerator generator =
        new UncorrelatedRandomVectorGenerator(new double[] { 50.0, 50.0 }, new double[] { 10.0, 10.0 },
                                              new GaussianRandomGenerator(g));
    MultivariateDifferentiableMultiStartOptimizer optimizer =
        new MultivariateDifferentiableMultiStartOptimizer(underlying, 10, generator);
    PointValuePair optimum =
        optimizer.optimize(200, circle, GoalType.MINIMIZE, new double[] { 98.680, 47.345 });
    Assert.assertEquals(200, optimizer.getMaxEvaluations());
    PointValuePair[] optima = optimizer.getOptima();
    for (PointValuePair o : optima) {
        Vector2D center = new Vector2D(o.getPointRef()[0], o.getPointRef()[1]);
        Assert.assertEquals(69.960161753, circle.getRadius(center), 1.0e-8);
        Assert.assertEquals(96.075902096, center.getX(), 1.0e-8);
        Assert.assertEquals(48.135167894, center.getY(), 1.0e-8);
    }
    Assert.assertTrue(optimizer.getEvaluations() > 70);
    Assert.assertTrue(optimizer.getEvaluations() < 90);
    Assert.assertEquals(3.1267527, optimum.getValue(), 1.0e-8);
}