Simple Neural-Network library for c++

Library is still under development

Features

This library supports easy to use and modify alorythyms.

Activation functions

Libray has some Transrer Functions built in, such as tanh, sigmoid, linear, reLU and softmax. But you are also able to create and add your own

Learning algorythyms

Library supports Back Propagation algorythym and critic learning.

Implementation

CMake

# include directoy
add_subdirectory(JimmyNeuron)

# link to your projsct
add_executable(project_name program.cpp)
target_link_libraries(project_name PUBLIC
    Jimmy
)

Simple program example

#include <iostream>
#include <JimmyNeuron/Jimmy.hpp>

int main(){
    // defying how many neurons will be craerted in each layer
    std::vector<int> topology{2,10,1}; 

    // creates Net instance
    Jimmy::Net mynet(topology, Jimmy::Methods::transFuncs::linear, Jimmy::Methods::lossFuncs::rmse, 0.01);

    // defying inputs, and desired outputs
    vector<double> inp = {0, 0};
    vector<double> prop = {0};

    // passing inputs and desired outputs for network to learn 
    mynet.feedForward(inp);
    mynet.backProp(prop);

    // get outputs n-th output neuron and loss of the net
    cout << "result:" << mynet.getResult(0) << "\tloss: " << mynet.getLoss() <<'\n';
}

Other examples

// create new type of TransferFunction
double foo(double weightedSum) //{ return some formula;}
double derivativeFoo(double weightedSum, double fooOutput) //{ return some derivative formula;}

//create instance of a new TransferFunctionM
const TransferFunction(foo, derivativeFoo);