public virtual void sensitivityTest()
{
//JAVA TO C# CONVERTER WARNING: The original Java variable was marked 'final':
//ORIGINAL LINE: final int nExamples = Y.length;
int nExamples = Y.Length;
//JAVA TO C# CONVERTER WARNING: The original Java variable was marked 'final':
//ORIGINAL LINE: final int n = XX.length;
int n = XX.Length;
//JAVA TO C# CONVERTER WARNING: The original Java variable was marked 'final':
//ORIGINAL LINE: final int nData = X.length;
int nData = X.Length;
for (int example = 0; example < nExamples; example++)
{
PiecewisePolynomialResultsWithSensitivity pp = PCHIP_S.interpolateWithSensitivity(X, Y[example]);
DoubleArray[] fdRes = fdSenseCal(X, Y[example], XX);
for (int i = 0; i < n; i++)
{
DoubleArray res = PPVAL_S.nodeSensitivity(pp, XX[i]);
for (int j = 0; j < nData; j++)
{
assertEquals("example: " + example + ", sample: " + i + ", node: " + j, fdRes[j].get(i), res.get(j), 1e-4);
}
}
}
}
/// <summary>
/// No clamped points added
/// </summary>
//JAVA TO C# CONVERTER TODO TASK: Most Java annotations will not have direct .NET equivalent attributes:
//ORIGINAL LINE: @Test public void notClampedTest()
public virtual void notClampedTest()
{
double[][] xValuesSet = new double[][]
{
new double[] { -5.0, -1.4, 3.2, 3.5, 7.6 },
new double[] { 1.0, 2.0, 4.5, 12.1, 14.2 },
new double[] { -5.2, -3.4, -3.2, -0.9, -0.2 }
};
double[][] yValuesSet = new double[][]
{
new double[] { -2.2, 1.1, 1.9, 2.3, -0.1 },
new double[] { 3.4, 5.2, 4.3, 1.1, 0.2 },
new double[] { 1.4, 2.2, 4.1, 1.9, 0.99 }
};
for (int k = 0; k < xValuesSet.Length; ++k)
{
double[] xValues = Arrays.copyOf(xValuesSet[k], xValuesSet[k].Length);
double[] yValues = Arrays.copyOf(yValuesSet[k], yValuesSet[k].Length);
int nData = xValues.Length;
double[] xyValues = new double[nData];
for (int j = 0; j < nData; ++j)
{
xyValues[j] = xValues[j] * yValues[j];
}
int nKeys = 100;
double interval = (xValues[nData - 1] - xValues[0]) / (nKeys - 1.0);
int n = INTERP.Length;
for (int i = 0; i < n; ++i)
{
ProductPiecewisePolynomialInterpolator interp = new ProductPiecewisePolynomialInterpolator(INTERP[i]);
for (int j = 0; j < nKeys; ++j)
{
double key = xValues[0] + interval * j;
InterpolatorTestUtil.assertRelative("notClampedTest", INTERP[i].interpolate(xValues, xyValues, key), interp.interpolate(xValues, yValues, key), EPS);
}
}
n = INTERP_SENSE.Length;
for (int i = 0; i < n; ++i)
{
ProductPiecewisePolynomialInterpolator interp = new ProductPiecewisePolynomialInterpolator(INTERP_SENSE[i]);
PiecewisePolynomialResultsWithSensitivity result = interp.interpolateWithSensitivity(xValues, yValues);
PiecewisePolynomialResultsWithSensitivity resultBase = INTERP_SENSE[i].interpolateWithSensitivity(xValues, xyValues);
for (int j = 0; j < nKeys; ++j)
{
double key = xValues[0] + interval * j;
InterpolatorTestUtil.assertRelative("notClampedTest", FUNC.evaluate(resultBase, key).get(0), FUNC.evaluate(result, key).get(0), EPS);
InterpolatorTestUtil.assertArrayRelative("notClampedTest", FUNC.nodeSensitivity(resultBase, key).toArray(), FUNC.nodeSensitivity(result, key).toArray(), EPS);
}
}
}
}
//-------------------------------------------------------------------------
private void testSensitivity(double[] xValues, double[] yValues, double[] keys, double delta)
{
PiecewisePolynomialWithSensitivityFunction1D func = new PiecewisePolynomialWithSensitivityFunction1D();
PiecewisePolynomialResultsWithSensitivity resultSensi = INTERP.interpolateWithSensitivity(xValues, yValues);
DoubleArray[] computedArray = func.nodeSensitivity(resultSensi, keys);
for (int i = 0; i < keys.Length; ++i)
{
double @base = func.evaluate(resultSensi, keys[i]).get(0);
DoubleArray computed = func.nodeSensitivity(resultSensi, keys[i]);
assertEquals(computed, computedArray[i]);
for (int j = 0; j < yValues.Length; ++j)
{
double[] yValuesBump = Arrays.copyOf(yValues, yValues.Length);
yValuesBump[j] += delta;
PiecewisePolynomialResult resultBump = INTERP.interpolate(xValues, yValuesBump);
double expected = (func.evaluate(resultBump, keys[i]).get(0) - @base) / delta;
assertEquals(computed.get(j), expected, delta);
}
}
}
/// <summary>
/// Left extrapolation by linear function unless extra node is added on the left
/// </summary>
private PiecewisePolynomialResult extrapolateByLinearFunction(PiecewisePolynomialResult result, double[] xValues)
{
int nIntervalsAll = result.NumberOfIntervals;
double[] nodes = result.Knots.toArray();
if (Math.Abs(xValues[xValues.Length - 1] - nodes[nIntervalsAll]) < EPS)
{
double lastNodeX = nodes[nIntervalsAll];
double lastNodeY = FUNC.evaluate(result, lastNodeX).get(0);
double extraNode = 2.0 * nodes[nIntervalsAll] - nodes[nIntervalsAll - 1];
double extraDerivative = FUNC.differentiate(result, lastNodeX).get(0);
double[] newKnots = new double[nIntervalsAll + 2];
Array.Copy(nodes, 0, newKnots, 0, nIntervalsAll + 1);
newKnots[nIntervalsAll + 1] = extraNode; // dummy node, outside the data range
double[][] newCoefMatrix = new double[nIntervalsAll + 1][];
for (int i = 0; i < nIntervalsAll; ++i)
{
newCoefMatrix[i] = Arrays.copyOf(result.CoefMatrix.row(i).toArray(), result.Order);
}
newCoefMatrix[nIntervalsAll] = new double[result.Order];
newCoefMatrix[nIntervalsAll][result.Order - 1] = lastNodeY;
newCoefMatrix[nIntervalsAll][result.Order - 2] = extraDerivative;
if (result is PiecewisePolynomialResultsWithSensitivity)
{
PiecewisePolynomialResultsWithSensitivity resultCast = (PiecewisePolynomialResultsWithSensitivity)result;
double[] extraSense = FUNC.nodeSensitivity(resultCast, lastNodeX).toArray();
double[] extraSenseDer = FUNC.differentiateNodeSensitivity(resultCast, lastNodeX).toArray();
DoubleMatrix[] newCoefSense = new DoubleMatrix[nIntervalsAll + 1];
for (int i = 0; i < nIntervalsAll; ++i)
{
newCoefSense[i] = resultCast.getCoefficientSensitivity(i);
}
//JAVA TO C# CONVERTER NOTE: The following call to the 'RectangularArrays' helper class reproduces the rectangular array initialization that is automatic in Java:
//ORIGINAL LINE: double[][] extraCoefSense = new double[resultCast.Order][extraSense.Length];
double[][] extraCoefSense = RectangularArrays.ReturnRectangularDoubleArray(resultCast.Order, extraSense.Length);
extraCoefSense[resultCast.Order - 1] = Arrays.copyOf(extraSense, extraSense.Length);
extraCoefSense[resultCast.Order - 2] = Arrays.copyOf(extraSenseDer, extraSenseDer.Length);
newCoefSense[nIntervalsAll] = DoubleMatrix.copyOf(extraCoefSense);
return(new PiecewisePolynomialResultsWithSensitivity(DoubleArray.copyOf(newKnots), DoubleMatrix.copyOf(newCoefMatrix), resultCast.Order, 1, newCoefSense));
}
return(new PiecewisePolynomialResult(DoubleArray.copyOf(newKnots), DoubleMatrix.copyOf(newCoefMatrix), result.Order, 1));
}
return(result);
}