Author: erans
Date: Wed Feb 16 22:48:43 2011
New Revision: 1071430
URL: http://svn.apache.org/viewvc?rev=1071430&view=rev
Log:
MATH-514
Added a "Parametric" inner class.
Modified:
commons/proper/math/trunk/src/main/java/org/apache/commons/math/analysis/function/Gaussian.java
commons/proper/math/trunk/src/test/java/org/apache/commons/math/analysis/function/GaussianTest.java
Modified:
commons/proper/math/trunk/src/main/java/org/apache/commons/math/analysis/function/Gaussian.java
URL:
http://svn.apache.org/viewvc/commons/proper/math/trunk/src/main/java/org/apache/commons/math/analysis/function/Gaussian.java?rev=1071430&r1=1071429&r2=1071430&view=diff
==============================================================================
---
commons/proper/math/trunk/src/main/java/org/apache/commons/math/analysis/function/Gaussian.java
(original)
+++
commons/proper/math/trunk/src/main/java/org/apache/commons/math/analysis/function/Gaussian.java
Wed Feb 16 22:48:43 2011
@@ -19,7 +19,10 @@ package org.apache.commons.math.analysis
import org.apache.commons.math.analysis.UnivariateRealFunction;
import org.apache.commons.math.analysis.DifferentiableUnivariateRealFunction;
+import org.apache.commons.math.analysis.ParametricUnivariateRealFunction;
import org.apache.commons.math.exception.NotStrictlyPositiveException;
+import org.apache.commons.math.exception.NullArgumentException;
+import org.apache.commons.math.exception.DimensionMismatchException;
import org.apache.commons.math.util.FastMath;
/**
@@ -78,8 +81,7 @@ public class Gaussian implements Differe
/** {@inheritDoc} */
public double value(double x) {
- final double diff = x - mean;
- return norm * FastMath.exp(-diff * diff * i2s2);
+ return value(x - mean, norm, i2s2);
}
/** {@inheritDoc} */
@@ -88,7 +90,7 @@ public class Gaussian implements Differe
/** {@inheritDoc} */
public double value(double x) {
final double diff = x - mean;
- final double g = Gaussian.this.value(x);
+ final double g = Gaussian.value(diff, norm, i2s2);
if (g == 0) {
// Avoid returning NaN in case of overflow.
@@ -99,4 +101,99 @@ public class Gaussian implements Differe
}
};
}
+
+ /**
+ * Parametric function where the input array contains the parameters of
+ * the Gaussian, ordered as follows:
+ * <ul>
+ * <li>Norm</li>
+ * <li>Mean</li>
+ * <li>Standard deviation</li>
+ * </ul>
+ */
+ public static class Parametric implements ParametricUnivariateRealFunction
{
+ /**
+ * Computes the value of the Gaussian at {@code x}.
+ *
+ * @param x Value for which the function must be computed.
+ * @param param Values of norm, mean and standard deviation.
+ * @return the value of the function.
+ * @throws NullArgumentException if {@code param} is {@code null}.
+ * @throws DimensionMismatchException if the size of {@code param} is
+ * not 3.
+ * @throws NotStrictlyPositiveException if {@code param[2]} is
negative.
+ */
+ public double value(double x,
+ double[] param) {
+ validateParameters(param);
+
+ final double diff = x - param[1];
+ final double i2s2 = 1 / (2 * param[2] * param[2]);
+ return Gaussian.value(diff, param[0], i2s2);
+ }
+
+ /**
+ * Computes the value of the gradient at {@code x}.
+ * The components of the gradient vector are the partial
+ * derivatives of the function with respect to each of the
+ * <em>parameters</em> (norm, mean and standard deviation).
+ *
+ * @param x Value at which the gradient must be computed.
+ * @param param Values of norm, mean and standard deviation.
+ * @return the gradient vector at {@code x}.
+ * @throws NullArgumentException if {@code param} is {@code null}.
+ * @throws DimensionMismatchException if the size of {@code param} is
+ * not 3.
+ * @throws NotStrictlyPositiveException if {@code param[2]} is
negative.
+ */
+ public double[] gradient(double x, double[] param) {
+ validateParameters(param);
+
+ final double norm = param[0];
+ final double diff = x - param[1];
+ final double sigma = param[2];
+ final double i2s2 = 1 / (2 * sigma * sigma);
+
+ final double n = Gaussian.value(diff, 1, i2s2);
+ final double m = norm * n * 2 * i2s2 * diff;
+ final double s = m * diff / sigma;
+
+ return new double[] { n, m, s };
+ }
+
+ /**
+ * Validates parameters to ensure they are appropriate for the
evaluation of
+ * the {@link #value(double,double[])} and {@link
#gradient(double,double[])}
+ * methods.
+ *
+ * @param param Values of norm, mean and standard deviation.
+ * @throws NullArgumentException if {@code param} is {@code null}.
+ * @throws DimensionMismatchException if the size of {@code param} is
+ * not 3.
+ * @throws NotStrictlyPositiveException if {@code param[2]} is
negative.
+ */
+ private void validateParameters(double[] param) {
+ if (param == null) {
+ throw new NullArgumentException();
+ }
+ if (param.length != 3) {
+ throw new DimensionMismatchException(param.length, 3);
+ }
+ if (param[2] <= 0) {
+ throw new NotStrictlyPositiveException(param[2]);
+ }
+ }
+ }
+
+ /**
+ * @param xMinusMean {@code x - mean}.
+ * @param norm Normalization factor.
+ * @param i2s2 Inverse of twice the square of the standard deviation.
+ * @return the value of the Gaussian at {@code x}.
+ */
+ private static double value(double xMinusMean,
+ double norm,
+ double i2s2) {
+ return norm * FastMath.exp(-xMinusMean * xMinusMean * i2s2);
+ }
}
Modified:
commons/proper/math/trunk/src/test/java/org/apache/commons/math/analysis/function/GaussianTest.java
URL:
http://svn.apache.org/viewvc/commons/proper/math/trunk/src/test/java/org/apache/commons/math/analysis/function/GaussianTest.java?rev=1071430&r1=1071429&r2=1071430&view=diff
==============================================================================
---
commons/proper/math/trunk/src/test/java/org/apache/commons/math/analysis/function/GaussianTest.java
(original)
+++
commons/proper/math/trunk/src/test/java/org/apache/commons/math/analysis/function/GaussianTest.java
Wed Feb 16 22:48:43 2011
@@ -19,6 +19,8 @@ package org.apache.commons.math.analysis
import org.apache.commons.math.analysis.UnivariateRealFunction;
import org.apache.commons.math.exception.NotStrictlyPositiveException;
+import org.apache.commons.math.exception.NullArgumentException;
+import org.apache.commons.math.exception.DimensionMismatchException;
import org.apache.commons.math.util.FastMath;
import org.junit.Assert;
@@ -84,4 +86,71 @@ public class GaussianTest {
Assert.assertTrue(Double.isNaN(dfdx.value(Double.NaN)));
}
+
+ @Test(expected=NullArgumentException.class)
+ public void testParametricUsage1() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.value(0, null);
+ }
+
+ @Test(expected=DimensionMismatchException.class)
+ public void testParametricUsage2() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.value(0, new double[] {0});
+ }
+
+ @Test(expected=NotStrictlyPositiveException.class)
+ public void testParametricUsage3() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.value(0, new double[] {0, 1, 0});
+ }
+
+ @Test(expected=NullArgumentException.class)
+ public void testParametricUsage4() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.gradient(0, null);
+ }
+
+ @Test(expected=DimensionMismatchException.class)
+ public void testParametricUsage5() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.gradient(0, new double[] {0});
+ }
+
+ @Test(expected=NotStrictlyPositiveException.class)
+ public void testParametricUsage6() {
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ g.gradient(0, new double[] {0, 1, 0});
+ }
+
+ @Test
+ public void testParametricValue() {
+ final double norm = 2;
+ final double mean = 3;
+ final double sigma = 4;
+ final Gaussian f = new Gaussian(norm, mean, sigma);
+
+ final Gaussian.Parametric g = new Gaussian.Parametric();
+ Assert.assertEquals(f.value(-1), g.value(-1, new double[] {norm, mean,
sigma}), 0);
+ Assert.assertEquals(f.value(0), g.value(0, new double[] {norm, mean,
sigma}), 0);
+ Assert.assertEquals(f.value(2), g.value(2, new double[] {norm, mean,
sigma}), 0);
+ }
+
+ @Test
+ public void testParametricGradient() {
+ final double norm = 2;
+ final double mean = 3;
+ final double sigma = 4;
+ final Gaussian.Parametric f = new Gaussian.Parametric();
+
+ final double x = 1;
+ final double[] grad = f.gradient(1, new double[] {norm, mean, sigma});
+ final double diff = x - mean;
+ final double n = FastMath.exp(-diff * diff / (2 * sigma * sigma));
+ Assert.assertEquals(n, grad[0], EPS);
+ final double m = norm * n * diff / (sigma * sigma);
+ Assert.assertEquals(m, grad[1], EPS);
+ final double s = m * diff / sigma;
+ Assert.assertEquals(s, grad[2], EPS);
+ }
}
