RichardBerg: MachineLearningHomework2

9/29/03

1) Invoke the random generator. Start an accumulator variable at 0 and an index at e(1). Increment the index through the e(i)'s and add p(i) to the variable. When the accumulator exceeds the random number, return e(i-1).

2) Run the previous algorithm as if the random number were 1, storing the intermediate values of the accumulator (distribution function) in a sorted container. When the new algorithm is called, search the container for the greatest value less than a random number and return its index.

3)

#include<iostream>

using namespace std;

#include "stdlib.h"
#include "time.h"

const int uniformSamples = 100;

double getNormalVal(double mean, double stddev)
{    
    // Central Limit Theorem:
    // The sum of N random #'s between 0 and 1 is normal @ N/2
    double randVal = 0;
    for (int i=0; i < uniformSamples; ++i)
    {
        randVal += rand() / double(RAND_MAX+1);
    }

    // Center around 0
    randVal -= uniformSamples / 2.;
    // Scale by std dev
    randVal *= stddev * 12./uniformSamples;
    // Shift to mean
    randVal += mean;

    return randVal;
}

int main(int argc, char ** argv)
{
    if (argc != 4)
    {
        cerr << "NormalSample mean stddev vals" << endl;
        return -1;
    }

    srand( time(NULL) );    

    double mean = strtod(argv[1], NULL);
    double stddev = strtod(argv[2], NULL);

    int vals = atoi(argv[3]);
    for (int i=0; i < vals; ++i)
    {
        cout << getNormalVal(mean, stddev) << endl;
    }

    return 0;
}

Back to MachineLearning

RichardBerg : MachineLearningHomework2