libbpp-core-dev 2.1.0-1 / usr / include / Bpp / Numeric / Random / RandomTools.h

//
// File RandomTools.h
// Author : Julien Dutheil
//          Sylvain Gaillard
// Last modification : Thu November 6 2008
//

/*
  Copyright or © or Copr. Bio++ Development Team, (November 17, 2004)

  This software is a computer program whose purpose is to provide classes
  for numerical calculus.

  This software is governed by the CeCILL  license under French law and
  abiding by the rules of distribution of free software.  You can  use, 
  modify and/ or redistribute the software under the terms of the CeCILL
  license as circulated by CEA, CNRS and INRIA at the following URL
  "http://www.cecill.info". 

  As a counterpart to the access to the source code and  rights to copy,
  modify and redistribute granted by the license, users are provided only
  with a limited warranty  and the software's author,  the holder of the
  economic rights,  and the successive licensors  have only  limited
  liability. 

  In this respect, the user's attention is drawn to the risks associated
  with loading,  using,  modifying and/or developing or reproducing the
  software by the user in light of its specific status of free software,
  that may mean  that it is complicated to manipulate,  and  that  also
  therefore means  that it is reserved for developers  and  experienced
  professionals having in-depth computer knowledge. Users are therefore
  encouraged to load and test the software's suitability as regards their
  requirements in conditions enabling the security of their systems and/or 
  data to be ensured and,  more generally, to use and operate it in the 
  same conditions as regards security. 

  The fact that you are presently reading this means that you have had
  knowledge of the CeCILL license and that you accept its terms.
*/

#ifndef _RANDOMTOOLS_H_
#define _RANDOMTOOLS_H_

#include "RandomFactory.h"
#include "../VectorExceptions.h"
#include "../VectorTools.h"
#include "../../Exceptions.h"

// From the STL:
#include <cmath>
#include <cassert>
#include <ctime>
#include <vector>

namespace bpp
{

  /**
   * @brief Utilitary function dealing with random numbers.
   *
   * This class uses Uniform01K generator by default.
   * It is possible to change this by setting the DEFAULT_GENERATOR variable.
   *
   * This class is adapted from Pupko's SEMPHY library.
   * It also borrow some code from Yang's PAML package.
   *
   * @see RandomFactory
   */
  class RandomTools
  {
  public:
    RandomTools() {}
    virtual ~RandomTools() {}

  private:
    /**
     *@brief functions for the computation of incompleteBeta
     */
    static double incompletebetafe(double a,
                                   double b,
                                   double x,
                                   double big,
                                   double biginv);
    static double incompletebetafe2(double a,
                                    double b,
                                    double x,
                                    double big,
                                    double biginv);
    static double incompletebetaps(double a,
                                   double b,
                                   double x,
                                   double maxgam);
  
  public:
    static RandomFactory* DEFAULT_GENERATOR;
    
    /**
     * @brief Get a double random value (between 0 and specified range).
     *
     * Note : the number you get is between 0 and entry not including entry !
     * @param entry Max number to reach.
     * @param generator Random number generator to use.
     */
    static double giveRandomNumberBetweenZeroAndEntry(double entry, const RandomFactory& generator = *DEFAULT_GENERATOR);

    /**
     * @brief Get a boolean random value.
     *
     * @param generator Random number generator to use.
     */
    static bool flipCoin(const RandomFactory& generator = *DEFAULT_GENERATOR);

    /**
     * @brief Get an integer random value (between 0 and specified range).
     *
     * Note : the number you get is between 0 and entry not including entry !
     * @param entry Max number to reach.
     * @param generator Random number generator to use.
     */
    static int giveIntRandomNumberBetweenZeroAndEntry(int entry, const RandomFactory& generator = *DEFAULT_GENERATOR);

    /**
     * @brief Set the default generator seed.
     *
     * @param seed New seed.
     */
    static void setSeed(long seed);

    /**
     * @return A random number drawn from a normal distribution.
     * @param mean The mean of the law.
     * @param variance The variance of the law.
     * @param generator The uniform generator to use.
     */
    static double randGaussian(double mean, double variance, const RandomFactory& generator = *DEFAULT_GENERATOR);
    
    /**
     * @return A random number drawn from a gamma distribution with unit scale (beta=1).
     * @param dblAlpha The alpha parameter.
     * @param generator The uniform generator to use.
     */
    static double randGamma(double dblAlpha, const RandomFactory& generator = *DEFAULT_GENERATOR);

    /**
     * @return A random number drawn from a gamma distribution.
     * @param alpha The alpha parameter.
     * @param beta The beta parameter.
     * @param generator The uniform generator to use.
     */
    static double randGamma(double alpha, double beta, const RandomFactory& generator = *DEFAULT_GENERATOR);
  
    /**
     * @return A random number drawn from a beta distribution.
     * @param alpha The alpha parameter.
     * @param beta The beta parameter.
     * @param generator The uniform generator to use.
     */

    static double randBeta(double alpha, double beta, const RandomFactory& generator = *DEFAULT_GENERATOR);
  
    /**
     * @return A random number drawn from an exponential distribution.
     * @param mean The mean of the distribution.
     * @param generator The uniform generator to use.
     */
    static double randExponential(double mean, const RandomFactory& generator = *DEFAULT_GENERATOR);

    /**
     * @brief Pick one element in a vector
     *
     * Pick one element randomly in a vector and return it.
     *
     * @param v The vector of elements.
     * @param replace If set to yes, then elements are allowed to be picked more than once, and therefore can be re-"placed" in the final sample.(default: false)
     * @return On element of the vector.
     * @throw EmptyVectorException if the vector is empty.
     *
     * @author Sylvain Gaillard
     */
    template<class T>
    static T pickOne(std::vector<T>& v, bool replace = false) throw (EmptyVectorException<T>) {
      if (v.empty())
        throw EmptyVectorException<T>("RandomTools::pickOne: input vector is empty", &v);
      size_t pos = static_cast<size_t>(RandomTools::giveIntRandomNumberBetweenZeroAndEntry(static_cast<int>(v.size())));
      if (replace)
        return v[pos];
      else {
        T e = v[pos];
        v[pos] = v.back();
        v.pop_back();
        return e;
      }
    }

    /**
     * @brief Sample a vector.
     *
     * The sample is a new vector of the specified size.
     * If the size of the sample is identical to the original vector,
     * the result is a shuffle of the original vector.
     *
     * @param vin The vector to sample.
     * @param vout [out] The output vector to fill, with the appropriate size.
     * @param replace Should sampling be with replacement?
     * @return A vector which is a sample of v.
     * @throw IndexOutOfBoundException if the sample size exceeds the original
     * size when sampling without replacement.
     * @throw EmptyVectorException if the vector is empty.
     *
     * @author Sylvain Gaillard
     */
    template<class T> 
    static void getSample(const std::vector<T>& vin, std::vector<T>& vout, bool replace = false) throw (EmptyVectorException<T>, IndexOutOfBoundsException)
    {
      if (vout.size() > vin.size() && !replace)
        throw IndexOutOfBoundsException("RandomTools::getSample: size exceeded v.size.", vout.size(), 0, vin.size());
      std::vector<size_t> hat(vin.size());
      for (size_t i = 0 ; i < vin.size() ; i++)
        hat[i] = i;
      for (size_t i = 0 ; i < vout.size() ; i++)
        vout[i] = vin[pickOne(hat, replace)];
    }

    /**
     * @brief Pick one element in a vector, with associated probability weights
     *
     * Pick one element randomly in a vector and return it.
     * If you choose to make the picking without replacement the element is
     * removed from the vector, and so is the corresponding weight
     *
     * @param v The vector of elements.
     * @param w The vector of weight associated to the v elements.
     * @param replace Should pick with replacement? (default: false)
     * @return On element of the vector.
     * @throw EmptyVectorException if the vector is empty.
     *
     * @author Julien Dutheil
     */
    template<class T>
    static T pickOne(std::vector<T>& v, std::vector<double>& w, bool replace = false) throw (EmptyVectorException<T>) {
      if (v.empty())
        throw EmptyVectorException<T>("RandomTools::pickOne (with weight): input vector is empty", &v);
      //Compute cumulative sum of weights:
      std::vector<double> sumw = VectorTools::cumSum(w);
      //Convert to cumulative distribution:
      sumw /= sumw.back();
      //Get random positions:
      double prob = RandomTools::giveRandomNumberBetweenZeroAndEntry(1.0);
      size_t pos = v.size() - 1;
      for (size_t i = 0; i < v.size(); ++i) {
        if (prob < sumw[i]) {
          pos = i;
          break;
        }
      }
      if (replace)
        return v[pos];
      else {
        T e = v[pos];
        v[pos] = v.back();
        v.pop_back();
        w[pos] = w.back();
        w.pop_back();
        return e;
      }
    }

    /**
     * @brief Sample a vector, with associated probability weights.
     *
     * The sample is a new vector of the specified size.
     * If the size of the sample is identical to the original vector,
     * the result is a shuffle of the original vector.
     *
     * It has to be noted that in case of sampling without replacement,
     * the effect of the weighting scheme will be lower as the sampe size becomes
     * close to the population size. In case the two are equal (pure permutations),
     * the weigths have no effect at all.
     *
     * @param vin The vector to sample.
     * @param w The vector of weights.
     * @param vout [out] The output vector to fill, with the appropriate size.
     * @param replace Should sampling be with replacement?
     * @return A vector which is a sample of v.
     * @throw IndexOutOfBoundException if the sample size exceeds the original
     * size when sampling without replacement.
     * @throw EmptyVectorException if the vector is empty.
     *
     * @author Julien Dutheil
     */
   template<class T> 
    static void getSample(const std::vector<T>& vin, const std::vector<double>& w, std::vector<T>& vout, bool replace = false) throw (EmptyVectorException<T>, IndexOutOfBoundsException)
    {
      if (vout.size() > vin.size() && !replace)
        throw IndexOutOfBoundsException("RandomTools::getSample (with weights): size exceeded v.size.", vout.size(), 0, vin.size());
      std::vector<size_t> hat(vin.size());
      for (size_t i = 0 ; i < vin.size() ; i++)
        hat[i] = i;
      std::vector<double> w2(w); //non const copy
      for (size_t i = 0 ; i < vout.size() ; i++)
        vout[i] = vin[pickOne(hat, w2, replace)];
    }

    /**
     * @brief Get a random state from a set of probabilities/scores.
     *
     * The input probabilities are scaled so that they sum to one.
     * If 'x' probabilities are provided as input, the output vector will contain values between 0 and 'x-1'.
     *
     * @param n The sample size.
     * @param probs The set of intput probabilities.
     * @return A vector of int values corresponding to the output states. States are supposed to be in the same order as the input probabilities, the first state being '0'.
     */ 
    static std::vector<size_t> randMultinomial(size_t n, const std::vector<double>& probs);

    /**
     * @name Probability functions.
     *
     * @{
     * Adapted from Yang's PAML package.
     *
     */

    /**
     * @brief Normal quantile function.
     *
     * Returns z so that Prob{x<z}=prob where x ~ N(0,1) and (1e-12)<prob<1-(1e-12)
     * returns (-9999) if in error
     * Odeh RE & Evans JO (1974) The percentage points of the normal distribution.
     * Applied Statistics 22: 96-97 (AS70)
     *
     * Newer methods:
     *  Wichura MJ (1988) Algorithm AS 241: the percentage points of the
     *    normal distribution.  37: 477-484.
     *  Beasley JD & Springer SG  (1977).  Algorithm AS 111: the percentage 
     *    points of the normal distribution.  26: 118-121.
     *
     * @param prob The probability.
     * @return The quantile corresponding to prob.
     */
    static double qNorm(double prob);
  
    /**
     * @brief Normal quantile function.
     *
     * Returns z so that Prob{x<z}=prob where x ~ N(mu,sigma^2) and (1e-12)<prob<1-(1e-12)
     * returns (-9999) if in error
     * Odeh RE & Evans JO (1974) The percentage points of the normal distribution.
     * Applied Statistics 22: 96-97 (AS70)
     *
     * Newer methods:
     *  Wichura MJ (1988) Algorithm AS 241: the percentage points of the
     *    normal distribution.  37: 477-484.
     *  Beasley JD & Springer SG  (1977).  Algorithm AS 111: the percentage 
     *    points of the normal distribution.  26: 118-121.
     *
     * @param prob The probability.
     * @param mu The mean of the distribution
     * @param sigma The standard deviation of the distribution
     * @return The quantile corresponding to prob.
     */
    static double qNorm(double prob, double mu, double sigma);

    /**
     * @brief Computes \f$ln\left(\Gamma\left(\alpha\right)\right)\f$ given \f$\alpha\f$.
     * 
     * Returns ln(gamma(alpha)) for alpha>0, accurate to 10 decimal places.  
     * Stirling's formula is used for the central polynomial part of the procedure.
     * Pike MC & Hill ID (1966) Algorithm 291: Logarithm of the gamma function.
     * Communications of the Association for Computing Machinery, 9:684
     *
     * @param alpha Alpha parameter.
     * @return \f$ln\left(\Gamma\left(\alpha\right)\right)\f$
     */
    static double lnGamma (double alpha);

    /**
     * @brief Returns the incomplete gamma ratio I(x,alpha).
     *
     * X is the upper limit of the integration and alpha is the shape parameter.
     * returns (-1) if in error
     * ln_gamma_alpha = ln(Gamma(alpha)), is almost redundant.
     * (1) series expansion     if (alpha>x || x<=1)
     * (2) continued fraction   otherwise
     * RATNEST FORTRAN by
     * Bhattacharjee GP (1970) The incomplete gamma integral.  Applied Statistics,
     * 19: 285-287 (AS32)
     *
     * @param x the upper limit of the integration.
     * @param alpha the shape parameter.
     * @param ln_gamma_alpha ln(Gamma(alpha)).
     */
    static double incompleteGamma(double x, double alpha, double ln_gamma_alpha);


    /**
     * @brief \f$\chi^2\f$ quantile function.
     * 
     * returns z so that Prob{x<z}=prob where x is Chi2 distributed with df=v
     * returns -1 if in error.   0.000002<prob<0.999998
     * RATNEST FORTRAN by
     * Best DJ & Roberts DE (1975) The percentage points of the 
     * Chi2 distribution.  Applied Statistics 24: 385-388.  (AS91)
     * Converted into C by Ziheng Yang, Oct. 1993.
     *
     * @param prob The probability.
     * @param v number of degree of freedom.
     * @return The quantile corresponding to prob.
     */
    static double qChisq(double prob, double v);

    /**
     * @brief \f$\chi^2\f$ cumulative probability function.
     *
     * @param x The quantile for which the probability should be computed.
     * @param v number of degree of freedom.
     * @return The corresponding probability of the quantile.
     */
    static double pChisq(double x, double v)
    {
      if (x < 0) return 0;
      return pGamma(x, v / 2, 0.5);
    }

    /**
     * @brief The Gamma quantile function.
     *
     * @param prob The probability.
     * @param alpha Alpha parameter.
     * @param beta  Beta parameter.
     * @return The quantile corresponding to prob.
     */
    static double qGamma(double prob, double alpha, double beta)
    {
      return qChisq(prob,2.0*(alpha))/(2.0*(beta));
    }

    /**
     * @brief \f$\Gamma\f$ cumulative probability function.
     *
     * @param x The quantile for which the probability should be computed.
     * @param alpha Alpha parameter.
     * @param beta  Beta parameter.
     * @return The corresponding probability of the quantile.
     * @throw Exception If alpha or beta is invalid (<0).
     *
     */
    static double pGamma(double x, double alpha, double beta) throw (Exception)
    {
      if (alpha < 0) throw Exception("RandomTools::pGamma. Negative alpha is not allowed.");
      if (beta < 0) throw Exception("RandomTools::pGamma. Negative beta is not allowed.");
      if (alpha == 0.) return 1.;
      return incompleteGamma(beta*x, alpha, lnGamma(alpha));
    }

    /** @} */
    
    /**
     * @name Other probability functions.
     *
     * Adapted from C routines for R programming langague
     *  Copyright (C) 1995, 1996  Robert Gentleman and Ross Ihaka
     *  Copyright (C) 1998    Ross Ihaka
     *  Copyright (C) 2000-2002 The R Development Core Team
     *  Copyright (C) 2003    The R Foundation

     * @{
     */
    /**
     * @brief Normal cumulative function.
     *
     * Returns Prob{x<=z} where x ~ N(0,1)
     
     * @param z the value.
     * @return The corresponding probability.
     */
    static double pNorm(double z);

    /* @brief Normal cumulative function.
    *
    * Returns Prob{x<=z} where x ~ N(mu,sigma^2)
     
    * @param z the value.
    * @param mu The mean of the distribution
    * @param sigma The standard deviation of the distribution
    * @return The corresponding probability.
    */
    
    static double pNorm(double z, double mu, double sigma);

    /**
     * @brief Computes
     * \f$ln\left(Beta\left(\alpha,\beta\right)\right)\f$ given
     * \f$\alpha\f$ and \f$b\eta\f$.
     * 
     * Returns ln(beta(alpha,beta)) for alpha>0 and beta>0.
     *
     * @param alpha, beta Alpha and Beta parameters.
     * @return \f$ln\left(Beta\left(\alpha,\beta\right)\right)\f$
     */

    static double lnBeta (double alpha, double beta);

    /**
     * @brief Returns the regularized incomplete beta function
     * @f$I_x(\alpha,\beta) = pbeta(x,\alpha,\beta@f$
     *
     * alpha and beta are the parameters of the function.
     *
     * Adapted From Cephes Math Library Release 2.8:  June, 2000
     * Copyright by Stephen L. Moshier
     * Under GPL License
     *
     * @param x the upper limit of the integration.
     * @param alpha, beta the shape parameters.
     */
    static double incompleteBeta(double x, double alpha, double beta);
    static double pBeta(double x, double alpha, double beta)
    {
      return incompleteBeta(x,alpha,beta);
    }

    /**
     * @brief The Beta quantile function.
     *
     * An adaptation from the C code of R
     *  Copyright (C) 1995, 1996  Robert Gentleman and Ross Ihaka
     *  Copyright (C) 1998--2007  The R Development Core Team
     *  based on code (C) 1979 and later Royal Statistical Society
     *
     * @param prob The probability.
     * @param alpha Alpha parameter.
     * @param beta  Beta parameter.
     * @return The quantile corresponding to prob.
     */
    static double qBeta(double prob, double alpha, double beta);

    /** @} */

  private:
    static double DblGammaGreaterThanOne(double dblAlpha, const RandomFactory& generator);
    static double DblGammaLessThanOne(double dblAlpha, const RandomFactory& generator);
  };

} //end of namespace bpp.

#endif  //_RANDOMTOOLS_H_