/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.commons.math3.distribution;
import org.apache.commons.math3.exception.NotStrictlyPositiveException;
import org.apache.commons.math3.exception.util.LocalizedFormats;
import org.apache.commons.math3.special.Gamma;
import org.apache.commons.math3.util.CombinatoricsUtils;
import org.apache.commons.math3.util.MathUtils;
import org.apache.commons.math3.util.FastMath;
import org.apache.commons.math3.random.RandomGenerator;
import org.apache.commons.math3.random.Well19937c;
/**
* Implementation of the Poisson distribution.
*
* @see <a href="http://en.wikipedia.org/wiki/Poisson_distribution">Poisson distribution (Wikipedia)</a>
* @see <a href="http://mathworld.wolfram.com/PoissonDistribution.html">Poisson distribution (MathWorld)</a>
*/
public class PoissonDistribution extends AbstractIntegerDistribution {
/**
* Default maximum number of iterations for cumulative probability calculations.
* @since 2.1
*/
public static final int DEFAULT_MAX_ITERATIONS = 10000000;
/**
* Default convergence criterion.
* @since 2.1
*/
public static final double DEFAULT_EPSILON = 1e-12;
/** Serializable version identifier. */
private static final long serialVersionUID = -3349935121172596109L;
/** Distribution used to compute normal approximation. */
private final NormalDistribution normal;
/** Distribution needed for the {@link #sample()} method. */
private final ExponentialDistribution exponential;
/** Mean of the distribution. */
private final double mean;
/**
* Maximum number of iterations for cumulative probability. Cumulative
* probabilities are estimated using either Lanczos series approximation
* of {@link Gamma#regularizedGammaP(double, double, double, int)}
* or continued fraction approximation of
* {@link Gamma#regularizedGammaQ(double, double, double, int)}.
*/
private final int maxIterations;
/** Convergence criterion for cumulative probability. */
private final double epsilon;
/**
* Creates a new Poisson distribution with specified mean.
*
* @param p the Poisson mean
* @throws NotStrictlyPositiveException if {@code p <= 0}.
*/
public PoissonDistribution(double p) throws NotStrictlyPositiveException {
this(p, DEFAULT_EPSILON, DEFAULT_MAX_ITERATIONS);
}
/**
* Creates a new Poisson distribution with specified mean, convergence
* criterion and maximum number of iterations.
*
* @param p Poisson mean.
* @param epsilon Convergence criterion for cumulative probabilities.
* @param maxIterations the maximum number of iterations for cumulative
* probabilities.
* @throws NotStrictlyPositiveException if {@code p <= 0}.
* @since 2.1
*/
public PoissonDistribution(double p, double epsilon, int maxIterations)
throws NotStrictlyPositiveException {
this(new Well19937c(), p, epsilon, maxIterations);
}
/**
* Creates a new Poisson distribution with specified mean, convergence
* criterion and maximum number of iterations.
*
* @param rng Random number generator.
* @param p Poisson mean.
* @param epsilon Convergence criterion for cumulative probabilities.
* @param maxIterations the maximum number of iterations for cumulative
* probabilities.
* @throws NotStrictlyPositiveException if {@code p <= 0}.
* @since 3.1
*/
public PoissonDistribution(RandomGenerator rng,
double p,
double epsilon,
int maxIterations)
throws NotStrictlyPositiveException {
super(rng);
if (p <= 0) {
throw new NotStrictlyPositiveException(LocalizedFormats.MEAN, p);
}
mean = p;
this.epsilon = epsilon;
this.maxIterations = maxIterations;
// Use the same RNG instance as the parent class.
normal = new NormalDistribution(rng, p, FastMath.sqrt(p),
NormalDistribution.DEFAULT_INVERSE_ABSOLUTE_ACCURACY);
exponential = new ExponentialDistribution(rng, 1,
ExponentialDistribution.DEFAULT_INVERSE_ABSOLUTE_ACCURACY);
}
/**
* Creates a new Poisson distribution with the specified mean and
* convergence criterion.
*
* @param p Poisson mean.
* @param epsilon Convergence criterion for cumulative probabilities.
* @throws NotStrictlyPositiveException if {@code p <= 0}.
* @since 2.1
*/
public PoissonDistribution(double p, double epsilon)
throws NotStrictlyPositiveException {
this(p, epsilon, DEFAULT_MAX_ITERATIONS);
}
/**
* Creates a new Poisson distribution with the specified mean and maximum
* number of iterations.
*
* @param p Poisson mean.
* @param maxIterations Maximum number of iterations for cumulative
* probabilities.
* @since 2.1
*/
public PoissonDistribution(double p, int maxIterations) {
this(p, DEFAULT_EPSILON, maxIterations);
}
/**
* Get the mean for the distribution.
*
* @return the mean for the distribution.
*/
public double getMean() {
return mean;
}
/** {@inheritDoc} */
public double probability(int x) {
final double logProbability = logProbability(x);
return logProbability == Double.NEGATIVE_INFINITY ? 0 : FastMath.exp(logProbability);
}
/** {@inheritDoc} */
@Override
public double logProbability(int x) {
double ret;
if (x < 0 || x == Integer.MAX_VALUE) {
ret = Double.NEGATIVE_INFINITY;
} else if (x == 0) {
ret = -mean;
} else {
ret = -SaddlePointExpansion.getStirlingError(x) -
SaddlePointExpansion.getDeviancePart(x, mean) -
0.5 * FastMath.log(MathUtils.TWO_PI) - 0.5 * FastMath.log(x);
}
return ret;
}
/** {@inheritDoc} */
public double cumulativeProbability(int x) {
if (x < 0) {
return 0;
}
if (x == Integer.MAX_VALUE) {
return 1;
}
return Gamma.regularizedGammaQ((double) x + 1, mean, epsilon,
maxIterations);
}
/**
* Calculates the Poisson distribution function using a normal
* approximation. The {@code N(mean, sqrt(mean))} distribution is used
* to approximate the Poisson distribution. The computation uses
* "half-correction" (evaluating the normal distribution function at
* {@code x + 0.5}).
*
* @param x Upper bound, inclusive.
* @return the distribution function value calculated using a normal
* approximation.
*/
public double normalApproximateProbability(int x) {
// calculate the probability using half-correction
return normal.cumulativeProbability(x + 0.5);
}
/**
* {@inheritDoc}
*
* For mean parameter {@code p}, the mean is {@code p}.
*/
public double getNumericalMean() {
return getMean();
}
/**
* {@inheritDoc}
*
* For mean parameter {@code p}, the variance is {@code p}.
*/
public double getNumericalVariance() {
return getMean();
}
/**
* {@inheritDoc}
*
* The lower bound of the support is always 0 no matter the mean parameter.
*
* @return lower bound of the support (always 0)
*/
public int getSupportLowerBound() {
return 0;
}
/**
* {@inheritDoc}
*
* The upper bound of the support is positive infinity,
* regardless of the parameter values. There is no integer infinity,
* so this method returns {@code Integer.MAX_VALUE}.
*
* @return upper bound of the support (always {@code Integer.MAX_VALUE} for
* positive infinity)
*/
public int getSupportUpperBound() {
return Integer.MAX_VALUE;
}
/**
* {@inheritDoc}
*
* The support of this distribution is connected.
*
* @return {@code true}
*/
public boolean isSupportConnected() {
return true;
}
/**
* {@inheritDoc}
* <p>
* <strong>Algorithm Description</strong>:
* <ul>
* <li>For small means, uses simulation of a Poisson process
* using Uniform deviates, as described
* <a href="http://irmi.epfl.ch/cmos/Pmmi/interactive/rng7.htm"> here</a>.
* The Poisson process (and hence value returned) is bounded by 1000 * mean.
* </li>
* <li>For large means, uses the rejection algorithm described in
* <quote>
* Devroye, Luc. (1981).<i>The Computer Generation of Poisson Random Variables</i>
* <strong>Computing</strong> vol. 26 pp. 197-207.
* </quote>
* </li>
* </ul>
* </p>
*
* @return a random value.
* @since 2.2
*/
@Override
public int sample() {
return (int) FastMath.min(nextPoisson(mean), Integer.MAX_VALUE);
}
/**
* @param meanPoisson Mean of the Poisson distribution.
* @return the next sample.
*/
private long nextPoisson(double meanPoisson) {
final double pivot = 40.0d;
if (meanPoisson < pivot) {
double p = FastMath.exp(-meanPoisson);
long n = 0;
double r = 1.0d;
double rnd = 1.0d;
while (n < 1000 * meanPoisson) {
rnd = random.nextDouble();
r *= rnd;
if (r >= p) {
n++;
} else {
return n;
}
}
return n;
} else {
final double lambda = FastMath.floor(meanPoisson);
final double lambdaFractional = meanPoisson - lambda;
final double logLambda = FastMath.log(lambda);
final double logLambdaFactorial = CombinatoricsUtils.factorialLog((int) lambda);
final long y2 = lambdaFractional < Double.MIN_VALUE ? 0 : nextPoisson(lambdaFractional);
final double delta = FastMath.sqrt(lambda * FastMath.log(32 * lambda / FastMath.PI + 1));
final double halfDelta = delta / 2;
final double twolpd = 2 * lambda + delta;
final double a1 = FastMath.sqrt(FastMath.PI * twolpd) * FastMath.exp(1 / (8 * lambda));
final double a2 = (twolpd / delta) * FastMath.exp(-delta * (1 + delta) / twolpd);
final double aSum = a1 + a2 + 1;
final double p1 = a1 / aSum;
final double p2 = a2 / aSum;
final double c1 = 1 / (8 * lambda);
double x = 0;
double y = 0;
double v = 0;
int a = 0;
double t = 0;
double qr = 0;
double qa = 0;
for (;;) {
final double u = random.nextDouble();
if (u <= p1) {
final double n = random.nextGaussian();
x = n * FastMath.sqrt(lambda + halfDelta) - 0.5d;
if (x > delta || x < -lambda) {
continue;
}
y = x < 0 ? FastMath.floor(x) : FastMath.ceil(x);
final double e = exponential.sample();
v = -e - (n * n / 2) + c1;
} else {
if (u > p1 + p2) {
y = lambda;
break;
} else {
x = delta + (twolpd / delta) * exponential.sample();
y = FastMath.ceil(x);
v = -exponential.sample() - delta * (x + 1) / twolpd;
}
}
a = x < 0 ? 1 : 0;
t = y * (y + 1) / (2 * lambda);
if (v < -t && a == 0) {
y = lambda + y;
break;
}
qr = t * ((2 * y + 1) / (6 * lambda) - 1);
qa = qr - (t * t) / (3 * (lambda + a * (y + 1)));
if (v < qa) {
y = lambda + y;
break;
}
if (v > qr) {
continue;
}
if (v < y * logLambda - CombinatoricsUtils.factorialLog((int) (y + lambda)) + logLambdaFactorial) {
y = lambda + y;
break;
}
}
return y2 + (long) y;
}
}
}