Examples of org.apache.commons.math3.analysis.differentiation.DerivativeStructure

Package org.apache.commons.math3.analysis.differentiation

Examples of org.apache.commons.math3.analysis.differentiation.DerivativeStructure

org.apache.commons.math3.analysis.differentiation.DerivativeStructure
american.edu/cas/mathstat/People/kalman/pdffiles/mmgautodiff.pdf">Doubly Recursive Multivariate Automatic Differentiation, Mathematics Magazine, vol. 75, no. 3, June 2002.
. Rall's numbers are an extension to the real numbers used throughout mathematical expressions; they hold the derivative together with the value of a function. Dan Kalman's derivative structures hold all partial derivatives up to any specified order, with respect to any number of free parameters. Rall's numbers therefore can be seen as derivative structures for order one derivative and one free parameter, and real numbers can be seen as derivative structures with zero order derivative and no free parameters.

{@link DerivativeStructure} instances can be used directly thanks tothe arithmetic operators to the mathematical functions provided as methods by this class (+, -, *, /, %, sin, cos ...).

Implementing complex expressions by hand using these classes is a tedious and error-prone task but has the advantage of having no limitation on the derivation order despite no requiring users to compute the derivatives by themselves. Implementing complex expression can also be done by developing computation code using standard primitive double values and to use {@link UnivariateFunctionDifferentiator differentiators} to create the {@link DerivativeStructure}-based instances. This method is simpler but may be limited in the accuracy and derivation orders and may be computationally intensive (this is typically the case for {@link FiniteDifferencesDifferentiator finite differencesdifferentiator}.

Instances of this class are guaranteed to be immutable.
@see DSCompiler @since 3.1

                        derivatives[i + 1] += gradient[j] * t[j].getPartialDerivative(orders);
                    }
                    orders[i] = 0;
                }


                return new DerivativeStructure(parameters, order, derivatives);


            }


        };
    }

View Full Code Here

                        final int n = x.length;


                        // delegate computation to underlying function
                        final DerivativeStructure[] dsX = new DerivativeStructure[n];
                        for (int i = 0; i < n; ++i) {
                            dsX[i] = new DerivativeStructure(n, 1, i, x[i]);
                        }
                        final DerivativeStructure[] y = f.value(dsX);


                        // extract Jacobian
                        final double[][] jacobian = new double[y.length][n];

View Full Code Here

                        for (int j = 0; j < n; ++j) {
                            derivatives[i + 1] += jacobian[k][j] * t[j].getPartialDerivative(orders);
                        }
                        orders[i] = 0;
                    }
                    merged[k] = new DerivativeStructure(parameters, order, derivatives);
                }


                return merged;


            }

View Full Code Here

        final double absoluteAccuracy = getAbsoluteAccuracy();


        double x0 = startValue;
        double x1;
        while (true) {
            final DerivativeStructure y0 = computeObjectiveValueAndDerivative(x0);
            x1 = x0 - (y0.getValue() / y0.getPartialDerivative(1));
            if (FastMath.abs(x1 - x0) <= absoluteAccuracy) {
                return x1;
            }


            x0 = x1;

View Full Code Here

        ++jacobianEvaluations;


        final DerivativeStructure[] dsPoint = new DerivativeStructure[params.length];
        final int nC = params.length;
        for (int i = 0; i < nC; ++i) {
            dsPoint[i] = new DerivativeStructure(nC, 1, i, params[i]);
        }
        final DerivativeStructure[] dsValue = jF.value(dsPoint);
        final int nR = getTarget().length;
        if (dsValue.length != nR) {
            throw new DimensionMismatchException(dsValue.length, nR);

View Full Code Here

            int i = 0;
            for (WeightedObservedPoint observed : observations) {


                // build the DerivativeStructure by adding first the value as a constant
                // and then adding derivatives
                DerivativeStructure vi = new DerivativeStructure(point.length, 1, f.value(observed.getX(), parameters));
                for (int k = 0; k < point.length; ++k) {
                    vi = vi.add(new DerivativeStructure(point.length, 1, k, 0.0));
                }


                values[i++] = vi;


            }

View Full Code Here

     * if the maximal number of evaluations is exceeded.
     */
    protected DerivativeStructure computeObjectiveValueAndDerivative(double point)
        throws TooManyEvaluationsException {
        incrementEvaluationCount();
        return function.value(new DerivativeStructure(1, 1, 0, point));
    }

View Full Code Here

    }


    @Override
    public DerivativeStructure[] value(DerivativeStructure[] variables) {
        DerivativeStructure[] f = new DerivativeStructure[m];
        DerivativeStructure sum = variables[0].getField().getZero();
        for (int i = 0; i < n; ++i) {
            sum = sum.add(variables[i].multiply(i + 1));
        }
        for (int i = 0; i < m; ++i) {
            f[i] = sum.multiply(i + 1).subtract(1);
        }
        return f;
    }

View Full Code Here

    }


    @Override
    public DerivativeStructure[] value(DerivativeStructure[] variables) {
        DerivativeStructure[] f = new DerivativeStructure[m];
        DerivativeStructure sum = variables[0].getField().getZero();
      for (int i = 1; i < (n - 1); ++i) {
          sum = sum.add(variables[i].multiply(i + 1));
      }
      for (int i = 0; i < (m - 1); ++i) {
        f[i] = sum.multiply(i).subtract(1);
      }
      f[m - 1] = variables[0].getField().getOne().negate();
      return f;
    }

View Full Code Here

      super(2, startParams, 0.0, buildArray(2, 1.0));
    }


    @Override
    public DerivativeStructure[] value(DerivativeStructure[] variables) {
        DerivativeStructure x1 = variables[0];
        DerivativeStructure x2 = variables[1];
        return new DerivativeStructure[] {
            x2.subtract(x1.multiply(x1)).multiply(10),
            x1.negate().add(1)
        };
    }

View Full Code Here

0 1 2 3 4 5 6 7 8 9

TOP

Related Classes of org.apache.commons.math3.analysis.differentiation.DerivativeStructure

org.apache.commons.math3.analysis.differentiation.MultivariateDifferentiableVectorFunction

org.apache.commons.math3.analysis.function.Constant

org.apache.commons.math3.analysis.function.GaussianTest

org.apache.commons.math3.analysis.function.LogitTest

org.apache.commons.math3.analysis.function.SigmoidTest

org.apache.commons.math3.analysis.function.Sin

org.apache.commons.math3.analysis.function.Sinc

org.apache.commons.math3.analysis.function.SincTest

org.apache.commons.math3.analysis.function.SqrtTest

org.apache.commons.math3.analysis.FunctionUtils

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.