Examples with PrimitiveMatrix org.ojalgo.matrix.PrimitiveMatrix used on opensource projects

Example 11 with PrimitiveMatrix

use of org.ojalgo.matrix.PrimitiveMatrix in project pyramid by cheng-li.

the class CBMInspector method covariance.

public static void covariance(CBM CBM, Vector vector, LabelTranslator labelTranslator) {
    int numClusters = CBM.getNumComponents();
    int numClasses = CBM.getNumClasses();
    double[] proportions = CBM.getMultiClassClassifier().predictClassProbs(vector);
    double[][] probabilities = new double[numClusters][numClasses];
    for (int k = 0; k < numClusters; k++) {
        for (int l = 0; l < numClasses; l++) {
            probabilities[k][l] = CBM.getBinaryClassifiers()[k][l].predictClassProb(vector, 1);
        }
    }
    // column vector
    DenseReceiver meanBuilder = factory.makeDense(numClasses, 1);
    for (int l = 0; l < numClasses; l++) {
        double sum = 0;
        for (int k = 0; k < numClusters; k++) {
            sum += proportions[k] * probabilities[k][l];
        }
        meanBuilder.set(l, 0, sum);
    }
    PrimitiveMatrix mean = meanBuilder.build();
    // System.out.println(mean);
    List<PrimitiveMatrix> mus = new ArrayList<>();
    for (int k = 0; k < numClusters; k++) {
        DenseReceiver muBuilder = factory.makeDense(numClasses, 1);
        for (int l = 0; l < numClasses; l++) {
            muBuilder.set(l, 0, probabilities[k][l]);
        }
        PrimitiveMatrix muK = muBuilder.build();
        mus.add(muK);
    }
    List<PrimitiveMatrix> sigmas = new ArrayList<>();
    for (int k = 0; k < numClusters; k++) {
        DenseReceiver sigmaBuilder = factory.makeDense(numClasses, numClasses);
        for (int l = 0; l < numClasses; l++) {
            double v = probabilities[k][l] * (1 - probabilities[k][l]);
            sigmaBuilder.set(l, l, v);
        }
        PrimitiveMatrix sigmaK = sigmaBuilder.build();
        sigmas.add(sigmaK);
    }
    PrimitiveMatrix covariance = factory.makeZero(numClasses, numClasses);
    for (int k = 0; k < numClusters; k++) {
        PrimitiveMatrix muk = mus.get(k);
        PrimitiveMatrix toadd = (sigmas.get(k).add(muk.multiply(muk.transpose()))).multiply(proportions[k]);
        covariance = covariance.add(toadd);
    }
    covariance = covariance.subtract(mean.multiply(mean.transpose()));
    // System.out.println("covariance = "+ Matrices.display(covariance));
    DenseReceiver correlationBuilder = factory.makeDense(numClasses, numClasses);
    for (int l = 0; l < numClasses; l++) {
        for (int j = 0; j < numClasses; j++) {
            double v = covariance.get(l, j).doubleValue() / (Math.sqrt(covariance.get(l, l).doubleValue()) * Math.sqrt(covariance.get(j, j).doubleValue()));
            correlationBuilder.set(l, j, v);
        }
    }
    PrimitiveMatrix correlation = correlationBuilder.build();
    // System.out.println("correlation = "+ Matrices.display(correlation));
    List<Pair<String, Double>> list = new ArrayList<>();
    for (int l = 0; l < numClasses; l++) {
        for (int j = 0; j < l; j++) {
            String s = "" + labelTranslator.toExtLabel(l) + ", " + labelTranslator.toExtLabel(j);
            double v = correlation.get(l, j).doubleValue();
            Pair<String, Double> pair = new Pair<>(s, v);
            list.add(pair);
        }
    }
    Comparator<Pair<String, Double>> comparator = Comparator.comparing(pair -> Math.abs(pair.getSecond()));
    List<Pair<String, Double>> top = list.stream().sorted(comparator.reversed()).limit(20).collect(Collectors.toList());
    System.out.println(top);
}

Example 12 with PrimitiveMatrix

use of org.ojalgo.matrix.PrimitiveMatrix in project ojAlgo by optimatika.

the class ConvexProblems method testP20080819.

/**
 * Another case of looping in the ActiveSetSolver's constraint (de)activation. Slightly different case (I
 * believe). The main reason/difficulty seemed to be that the algorithm would both add and remove
 * constraints in the iteration. Modified the algorithm to only do one thing with each iteration - either
 * add or remove.
 */
@SuppressWarnings("unchecked")
@Test
public void testP20080819() {
    final Factory<PrimitiveMatrix> tmpMtrxFact = PrimitiveMatrix.FACTORY;
    final NumberContext tmpEvalCntxt = StandardType.DECIMAL_032;
    final BasicMatrix[] tmpMatrices = new PrimitiveMatrix[8];
    tmpMatrices[0] = tmpMtrxFact.rows(new double[][] { { 1.0, 1.0, 1.0, 1.0 } });
    tmpMatrices[1] = tmpMtrxFact.rows(new double[][] { { 1.0 } });
    tmpMatrices[2] = tmpMtrxFact.rows(new double[][] { { 15.889978159746546, 7.506345724913546, 0.8416674706550127, 0.435643236753381 }, { 7.506345724913546, 8.325860065234632, 0.4230651628792374, 0.1670802923999648 }, { 0.8416674706550127, 0.4230651628792374, 1.00134099479915, 0.6558469727234849 }, { 0.435643236753381, 0.1670802923999648, 0.6558469727234849, 0.6420451103682865 } });
    tmpMatrices[3] = tmpMtrxFact.rows(new double[][] { { -0.15804736429388952 }, { -0.11226063792731895 }, { -0.10509261785657838 }, { -0.0848686735786316 } });
    tmpMatrices[4] = tmpMtrxFact.rows(new double[][] { { 1.0, 0.0, 0.0, 0.0 }, { 0.0, 1.0, 0.0, 0.0 }, { 0.0, 0.0, 1.0, 0.0 }, { 0.0, 0.0, 0.0, 1.0 }, { -0.15804736429388952, -0.11226063792731895, -0.10509261785657838, -0.0848686735786316 }, { -1.0, 0.0, 0.0, 0.0 }, { 0.0, -1.0, 0.0, 0.0 }, { 0.0, 0.0, -1.0, 0.0 }, { 0.0, 0.0, 0.0, -1.0 } });
    tmpMatrices[5] = tmpMtrxFact.rows(new double[][] { { 0.9 }, { 0.8 }, { 0.7 }, { 0.6 }, { 0.0 }, { -0.1 }, { -0.2 }, { -0.3 }, { -0.4 } });
    tmpMatrices[6] = tmpMtrxFact.rows(new double[][] { { 0.1 }, { 0.2 }, { 0.3 }, { 0.4 } });
    tmpMatrices[7] = null;
    final MatrixStore<Double>[] retVal = new MatrixStore[tmpMatrices.length];
    for (int i = 0; i < retVal.length; i++) {
        if (tmpMatrices[i] != null) {
            if (i == 3) {
                retVal[i] = PrimitiveDenseStore.FACTORY.copy(tmpMatrices[i].negate());
            } else {
                retVal[i] = PrimitiveDenseStore.FACTORY.copy(tmpMatrices[i]);
            }
        }
    }
    final ConvexSolver.Builder tmpBuilder = new ConvexSolver.Builder(retVal);
    // final ActiveSetSolver tmpSolver = new ActiveSetSolver(tmpMatrices);
    final ConvexSolver tmpSolver = tmpBuilder.build();
    // Test that the matrices were input in the right order
    // JUnitUtils.assertEquals(tmpSolver.getAE(), tmpMatrices[0].toPrimitiveStore(),
    // tmpEvalCntxt);
    // JUnitUtils.assertEquals(tmpSolver.getBE(), tmpMatrices[1].toPrimitiveStore(),
    // tmpEvalCntxt);
    // JUnitUtils.assertEquals(tmpSolver.getQ(), tmpMatrices[2].toPrimitiveStore(),
    // tmpEvalCntxt);
    // JUnitUtils.assertEquals(tmpSolver.getC(), tmpMatrices[3].negate().toPrimitiveStore(),
    // tmpEvalCntxt);
    // JUnitUtils.assertEquals(tmpSolver.getAI(), tmpMatrices[4].toPrimitiveStore(),
    // tmpEvalCntxt);
    // JUnitUtils.assertEquals(tmpSolver.getBI(), tmpMatrices[5].toPrimitiveStore(),
    // tmpEvalCntxt);
    final Optimisation.Result tmpResult = tmpSolver.solve();
    TestUtils.assertEquals(tmpMatrices[6], RationalMatrix.FACTORY.columns(tmpResult), tmpEvalCntxt);
    OptimisationConvexTests.assertDirectAndIterativeEquals(tmpBuilder, null);
}

Example 13 with PrimitiveMatrix

use of org.ojalgo.matrix.PrimitiveMatrix in project ojAlgo-finance by optimatika.

the class PortfolioProblems method testP20110614.

/**
 * A user claimed he got constraint breaking weights using these figures.
 */
@Test
public void testP20110614() {
    final Builder<PrimitiveMatrix> tmpCovarsBuilder = PrimitiveMatrix.FACTORY.getBuilder(3, 3);
    tmpCovarsBuilder.set(0, 0, 0.04);
    tmpCovarsBuilder.set(0, 1, 0.01);
    tmpCovarsBuilder.set(0, 2, 0.02);
    tmpCovarsBuilder.set(1, 0, 0.01);
    tmpCovarsBuilder.set(1, 1, 0.09);
    tmpCovarsBuilder.set(1, 2, 0.01);
    tmpCovarsBuilder.set(2, 0, 0.02);
    tmpCovarsBuilder.set(2, 1, 0.01);
    tmpCovarsBuilder.set(2, 2, 0.16);
    final BasicMatrix tmpCovars = tmpCovarsBuilder.build();
    final Builder<PrimitiveMatrix> tmpReturnsBuilder = PrimitiveMatrix.FACTORY.getBuilder(3, 1);
    tmpReturnsBuilder.set(0, 0, 0.10);
    tmpReturnsBuilder.set(1, 0, 0.15);
    tmpReturnsBuilder.set(2, 0, 0.18);
    final BasicMatrix tmpReturs = tmpReturnsBuilder.build();
    final MarketEquilibrium tmpME = new MarketEquilibrium(tmpCovars);
    final MarkowitzModel tmpMarkowitz = new MarkowitzModel(tmpME, tmpReturs);
    for (int i = 1; i < 10; i++) {
        tmpMarkowitz.setRiskAversion(new BigDecimal(i));
        final List<BigDecimal> tmpWeights = tmpMarkowitz.getWeights();
        for (final BigDecimal tmpBigDecimal : tmpWeights) {
            if ((tmpBigDecimal.compareTo(BigMath.ZERO) == -1) || (tmpBigDecimal.compareTo(BigMath.ONE) == 1)) {
                TestUtils.fail("!(0.0 <= " + tmpBigDecimal + " <= 1.0)");
            }
        }
    }
    // As the Markowitz model built it the problem
    final MatrixStore<Double> tmpQ = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 4.0, 1.0, 2.0 }, { 1.0, 9.0, 1.0 }, { 2.0, 1.0, 16.0 } });
    final MatrixStore<Double> tmpC = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 10.0 }, { 15.0 }, { 18.0 } });
    final MatrixStore<Double> tmpAE = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 1.0, 1.0, 1.0 } });
    final MatrixStore<Double> tmpBE = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 1.0 } });
    MatrixStore<Double> tmpAI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { -1.0, 0.0, 0.0 }, { 0.0, -1.0, 0.0 }, { 0.0, 0.0, -1.0 } });
    MatrixStore<Double> tmpBI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 0.0 }, { 0.0 }, { 0.0 } });
    final MatrixStore<Double> tmpX = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 0.0 }, { 0.5217391304347826 }, { 0.4782608695652173 } });
    ConvexSolver.Builder tmpBuilder = new ConvexSolver.Builder(tmpQ, tmpC).equalities(tmpAE, tmpBE).inequalities(tmpAI, tmpBI);
    ConvexSolver tmpSolver = tmpBuilder.build();
    // tmpSolver.options.debug(ConvexSolver.class);
    Optimisation.Result tmpResult = tmpSolver.solve();
    // BasicMatrix tmpSolution = tmpResult.getSolution();
    TestUtils.assertEquals(tmpX, tmpResult, new NumberContext(7, 6));
    // As (I believe) the user built it
    // 
    // The only *problem* I found was that he did not set lower limits
    // on the portfolio weights, which you have to do. No problem with
    // ojAlgo.
    tmpAI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 1.0, 0.0, 0.0 }, { 0.0, 1.0, 0.0 }, { 0.0, 0.0, 1.0 } });
    tmpBI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 1.0 }, { 1.0 }, { 1.0 } });
    tmpBuilder = new ConvexSolver.Builder(tmpQ, tmpC).equalities(tmpAE, tmpBE).inequalities(tmpAI, tmpBI);
    tmpSolver = tmpBuilder.build();
    tmpResult = tmpSolver.solve();
    // Should NOT be equal in this case!
    TestUtils.assertFalse(Access1D.equals(tmpX, tmpResult, new NumberContext(7, 6)));
    // No problem with both the lower and upper limits set.
    tmpAI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { -1.0, 0.0, 0.0 }, { 0.0, -1.0, 0.0 }, { 0.0, 0.0, -1.0 }, { 1.0, 0.0, 0.0 }, { 0.0, 1.0, 0.0 }, { 0.0, 0.0, 1.0 } });
    tmpBI = PrimitiveDenseStore.FACTORY.rows(new double[][] { { 0.0 }, { 0.0 }, { 0.0 }, { 1.0 }, { 1.0 }, { 1.0 } });
    tmpBuilder = new ConvexSolver.Builder(tmpQ, tmpC).equalities(tmpAE, tmpBE).inequalities(tmpAI, tmpBI);
    tmpSolver = tmpBuilder.build();
    tmpResult = tmpSolver.solve();
    TestUtils.assertEquals(tmpX, tmpResult, new NumberContext(7, 6));
}

Example 14 with PrimitiveMatrix

use of org.ojalgo.matrix.PrimitiveMatrix in project ojAlgo-finance by optimatika.

the class PortfolioProblems method testP20170508.

@Test
public void testP20170508() {
    Builder<PrimitiveMatrix> tmpBuilder = PrimitiveMatrix.FACTORY.getBuilder(2, 2);
    tmpBuilder.add(0, 0, 0.040000);
    tmpBuilder.add(0, 1, 0.1000);
    tmpBuilder.add(1, 0, 0.1000);
    tmpBuilder.add(1, 1, 0.250000);
    final BasicMatrix covariances = tmpBuilder.build();
    tmpBuilder = PrimitiveMatrix.FACTORY.getBuilder(2);
    tmpBuilder.add(0, 0.20000);
    tmpBuilder.add(1, 0.40000);
    final BasicMatrix returns = tmpBuilder.build();
    final MarketEquilibrium marketEq = new MarketEquilibrium(covariances);
    final MarkowitzModel markowitzModel = new MarkowitzModel(marketEq, returns);
    for (int r = 0; r <= 10; r++) {
        final BigDecimal targetReturn = StandardType.PERCENT.enforce(new BigDecimal(0.2 + (0.02 * r)));
        markowitzModel.setTargetReturn(targetReturn);
        markowitzModel.optimiser().validate(false);
        markowitzModel.optimiser().debug(false);
        final List<BigDecimal> tmpWeights = markowitzModel.getWeights();
        if (DEBUG) {
            BasicLogger.debug("{} => {} {}", targetReturn, markowitzModel.optimiser().getState(), markowitzModel.toSimplePortfolio());
        }
        TestUtils.assertTrue("Optimiser completed normally", markowitzModel.optimiser().getState().isOptimal());
        TestUtils.assertTrue("Weights sum to 100%", tmpWeights.get(0).add(tmpWeights.get(1)).setScale(2, RoundingMode.HALF_EVEN).compareTo(BigMath.ONE) == 0);
        TestUtils.assertEquals("Return is close to target", targetReturn, markowitzModel.getMeanReturn(), StandardType.PERCENT.newPrecision(2).newScale(2));
    }
}

Example 15 with PrimitiveMatrix

use of org.ojalgo.matrix.PrimitiveMatrix in project ojAlgo-finance by optimatika.

the class FinanceUtils method makeCovarianceMatrix.

/**
 * @param timeSeriesCollection
 * @return Annualised covariances
 */
public static <V extends Number> BasicMatrix makeCovarianceMatrix(final Collection<CalendarDateSeries<V>> timeSeriesCollection) {
    final CoordinationSet<V> tmpCoordinator = new CoordinationSet<>(timeSeriesCollection).prune();
    final ArrayList<SampleSet> tmpSampleSets = new ArrayList<>();
    for (final CalendarDateSeries<V> tmpTimeSeries : timeSeriesCollection) {
        final double[] values = tmpCoordinator.get(tmpTimeSeries.getName()).asPrimitive().toRawCopy1D();
        final int tmpSize1 = values.length - 1;
        final double[] retVal = new double[tmpSize1];
        for (int i = 0; i < tmpSize1; i++) {
            retVal[i] = PrimitiveFunction.LOG.invoke(values[i + 1] / values[i]);
        }
        final SampleSet tmpMakeUsingLogarithmicChanges = SampleSet.wrap(Access1D.wrap(retVal));
        tmpSampleSets.add(tmpMakeUsingLogarithmicChanges);
    }
    final int tmpSize = timeSeriesCollection.size();
    final Builder<PrimitiveMatrix> retValStore = PrimitiveMatrix.FACTORY.getBuilder(tmpSize, tmpSize);
    final double tmpToYearFactor = (double) CalendarDateUnit.YEAR.size() / (double) tmpCoordinator.getResolution().size();
    SampleSet tmpRowSet;
    SampleSet tmpColSet;
    for (int j = 0; j < tmpSize; j++) {
        tmpColSet = tmpSampleSets.get(j);
        for (int i = 0; i < tmpSize; i++) {
            tmpRowSet = tmpSampleSets.get(i);
            retValStore.set(i, j, tmpToYearFactor * tmpRowSet.getCovariance(tmpColSet));
        }
    }
    return retValStore.build();
}