lolkat247 / nuron_network

nuron_network

Project aimed at allowing users to create custom nural networks in C++ without the hassle of learning calculus.

Getting Started

These instructions will get you a copy of the headers and the functions inside them.

Prerequisites

Things you should have:

• General C++ knowlege
• patience

Explanation of functions

Sigmoid

Used for probability.

var2 = sigmoid(var1);

Sigmoid Prime

Used as internal function.

var2 = sigmoid_p(var1);

Randint

Random integer between 0 and var1.

var2 = randint(var1);

fexists

Check if a file exists for not, returns 1 on success.

if (fexists("file.txt")){
  //Some code
};

nuron_tr?

Used to generate algorithm from training data.

• tr2 uses 2 weights,
• tr3 uses 3 weights,
• etc...

This function will create a file named with the string you provided + .txt (text file). YOU MUST EDIT THIS DATA IN ORDER FOR IT TO WORK! The data file created MUST have an empty newline at the end or backp will not work. The function nuron_tr? will return 0 on success and 1 on error.

Currently supports 2-3

float w1, w2, b
if (nuron_tr3("nuron1", w1, w2, b, true){ // True is optional, it allows debug output
  cout << "Algorithm not generated." << endl;
}
else {
  // next bit of code
};

determine_?

Used to generate probable output based on algorithm and outputs to provided variables (float type). Output will be: 0 < output < 1.

• _2 uses 2 weights,
• _3 uses 3 weights,
• etc...

float var1 = 1 // test points
float var2 = 1

float w1 = 0.5 // weights and bias
float w2 = 0.5
float b = 0.5

cout << determine_2(var1, var2, w1, w2, b) << endl;

backp

Used to back propagate data into the training file.

string new_data_line = "2;4;1;0";
backp("nuron1", new_data_line);

Deployment

Download newest version and #include all modules you'd like to use. All nuron_3.h or higher files depend on nuron_2.h.

Example

Test code:

main.cpp

#include "nuron_2.h"
#include "nuron_3.h"
using namespace std;

int main(){
    float aw1, aw2, aw3, ab, ar;
    nuron_tr3("nuron_1-1", aw1, aw2, aw3, ab, true);

    float var1, var2, var3;
    cout << "var1: ";
    cin >> var1;
    cout << "var2: ";
    cin >> var2;
    cout << "var3: ";
    cin >> var3;

    ar = determine_3(var1, var2, var3, aw1, aw2, aw3, ab);
    cout << ar << endl;

    return 0;
};

nuron_1-1.txt

0;0;0
4;4;0
6;6;1
9;9;1

First Run

Output:

nuron_1-1: No data file to load, making new file...done
nuron_1-1: Please exit and edit data file: nuron_1-1.txt accordingly.
var1: (cin)1
var2: (cin)1
var3: (cin)1
0.500000

Second Run

Output

nuron_1-1: File acknowledged.
nuron_1-1: File loaded.
nuron_1-1: 1s to train
var1: (cin)1
var2: (cin)1
var3: (cin)1
0.000048

Authors

• Jacob Iglesias - Initial work - lolkat247

License

This project is licensed under the GNU v3 License - see the LICENSE.md file for details.

Acknowledgments

• giant_nural_network
• My friends over at Stack Overflow
• Hat tip to anyone whose code was used.