KNearest Neighbour Algorithm

This is an implementation of the KNearest Neighbour algorithm in C++, this repo reads data from the MNIST dataset. I added a MNIST file parser to read image data and labels. The image data is wrapped in an Image object where the pixel data is stored in a Eigen Matrix to allow for effiecent calculation of the euclidian distance when using the algorithm.

Caution this project is dependent on the Eigen Library

Type make to create the executable binary and to run the program

$ make

Example Main.cpp

#include "../include/DataHandler.hpp"
#include "../include/KNNAlgo.hpp"
#include <iostream>

using namespace std;

int main(){

    DataHandler dh;
    // False to specify that it is training dataset
    dh.prepareData("path_image_file", "path_label_file",false);
    // True to indicate that data is test dataset
    dh.prepareData("path_image_file_test", "path_label_file_test",true);

    double count = 0;
    double iterNum = 10;
    KNNAlgo knn;

    for(int i = 0; i < iterNum; i++){
        bool eval = knn.Classify(dh.get_Image_Data(), dh.get_Test_Image_Data()[i]);
        if(eval) count++;
    }

    cout << "Accuracy: " << (count/iterNum*100) << "%\n";

    return 0;
}

What I learned

• KNearest Neighbour Algorithm
• Parsing MNIST file
• Eigen Matrix Library