[프로그래밍] 각종 난수 발생 함수 링크

일반적으로 사용하는 rand() 는 uniform한 난수를 발생시킨다.

normal distribution이나 exponential distribution ... 등이 필요하면 아래 사이트에서 참고하여 작성하면 되겠다.

http://www.pamvotis.org/vassis/RandGen.htm

#include <stdio.h>

#include <stdlib.h>

#include <time.h>

#include <math.h>

 

 

//Note: float variables are used. If you want to generate integer variables, just change the

//return type of the function and all other types which are related to the result...

float Constant(float mean);

float Uniform(float mean);

float Exponential(float mean);

int   Poisson(float mean);

float Pareto(float alpha);

float Normal(float mean, float stdev);

int    Geometric(float p);

float Weibull(float scale, float shape);

float Erlang(int scale, float shape);

 

void main(void)

{

            float mean;

 

            //Seed is the parameter of srand (the random number generator function of C).

            //srand takes an integer as a parameter. If this integer is the same,

            //then each time the program is exucuted, the same random numbers

            //will be generated. In order to avoid this, we put the current time as the parameter of

            //the srand function.

            int seed = time(NULL);

           

            //Initiate the random number generator. The initiation must be done outside the generation

            //functions, and usually is done only once.

            srand(seed);

           

            printf("Give Mean Value:");

            scanf("%f",&mean);

           

            printf("\nConstant: %f\n",Constant(mean));

            printf("Uniform: %f\n",Uniform(mean));

            printf("Exponential: %f\n",Exponential(mean));

            printf("Poisson: %d\n",Poisson(mean));

            printf("\nMore results can be added for other functions too...\n\n");

}

 

 

float Constant (float mean)

{

            return mean;

}

 

 

float Uniform(float mean)

{

            //Generate a random number between 0 and 1.

            //REMEMBER: Always cast the oparands of a division to float, or truncation will be performed.

            float R;

            R = (float)rand()/(float)(RAND_MAX+1);

           

            return  2*mean*R;

}

 

 

float Exponential(float mean)

{

            //Generate a random number between 0 and 1.

            //REMEMBER: Always cast the oparands of a division to float, or truncation will be performed.

            float R;

            R = (float)rand()/(float)(RAND_MAX+1);

           

            //Put it as a parameter to the inverse distribution function.

            return  -mean*log(R);

}

 

 

int Poisson(float mean) //Special technique required: Box-Muller method...

{

            float R;

            float sum = 0;

            int i;

            i=-1;

            float z;

 

            while(sum <=mean)

            {

 

                        R = (float)rand()/(float)(RAND_MAX+1);

                        z = -log(R);

                        sum+= z;

                        i++;

            }

            return i;

}

                       

 

float Pareto(float alpha)

{

 

            float R;

            R = (float)rand()/(float)(RAND_MAX+1);

            return (float)1/(float)(pow(R,(float)1/alpha));

}

 

 

float Normal(float mean, float stdev)

{

 

            float R1;

            R1 = (float)rand()/(float)(RAND_MAX+1);

            float R2;

            R2 = (float)rand()/(float)(RAND_MAX+1);

            return mean + stdev*cos(2*3.14*R1)*sqrt(-log(R2));

}

 

 

int Geometric(float p)

{

            float R;

            R= (float)rand()/(float)(RAND_MAX+1);

            return (int)(log(R)/log(1-p) - 1);

}

 

 

float Weibull(float scale, float shape)

{

            float R;

            R= (float)rand()/(float)(RAND_MAX+1);

            return scale*pow(-log(R),(float)1/shape);

}

 

 

float Erlang(int scale, float shape)          //Special technique required...

{

            float R=1;

            int i;

            for (i=1;i<=scale;i++)

                        R = R*((float)rand()/(float)(RAND_MAX+1));

            return -shape*log(R);

}