Do you want to develop your of artificial intelligence application from the scratch ? Want to learn how you can develop a simple artificial neuron model in C++ with Classes ? In this post we will explain with a very simple artificial neuron example using with Classes. We have released A Simple Artificial Neuron Model in C++ and Very Simple Artificial Neural Network (ANN) Example in C++ and we also released Array Based Simple Artificial Neuron Model in C++ before. We highly recommend reading Introduction to AI Technologies and them if you are new to AI technology.
A Simple AI Model
We have presented this Simple AI neuron before, let’s remember. A Minimum Artificial Neuron has an activation value (a), an activation function ( phi() ) and weighted (w) input net links. So it has one activation value, one activation function and one or more weights depend on the number of its input nets.
This is a very simple artificial neural network in AI technology . Now let’s go with this example and improve our neuron models and lets create an artificial neuron model by using arrays.
Class Based Artificial Neuron Model in C++
This is another simple neuron example with classes. This is a modern way to create a neuron in C++ than using structs or arrays. This example is also good for modern C++ applications. We have explained Classes very detailed in our LearnCPlusPlus.org web site. If you are new to Classes and Objects first please read Learn Classes and Objects in C++ and you will find a lot of Class related topics that teach different usage and methods in Classes.
Object Oriented Programming is a way to integrate with objects which can contain data in the form (attributes or properties of objects), and code blocks in the form of procedures (methods, functions of objects). Classes are defined in C++ using keyword class followed by the name of the class. Classes are the blueprint for the objects and they are user-defined data types that we can use in our program, and they work as an object constructor. Objects are an instantiation of a class, In another term. In C++ programming, most of the commands are associated with classes and objects, along with their attributes and methods.
If you have a complex network, this call based artificial neuron model is more safe and functional to calculate activation of neurons. Structs and arrays are easy to allocate memory and easy to use on static neural networks but Classes are the best if you are strongly know their all features.
Let’s create our simple Tneuron class (T means it is a Type),
|
1 2 3 4 5 6 7 8 |
class Tneuron // neuron class { public: float a; // activaty of each neurons float w[NN+1]; // weight of links between each neurons } |
In this Tneuron class, we can create a constructor with its same name Tneuron() as below,
|
1 2 3 4 5 6 7 |
Tneuron() { a=0; for(int i=0; i<=NN; i++) w[i]=-1; // if weight is negative there is no link } |
In this Tneuron class, we can also define a linear activation function as below,
|
1 2 3 4 5 6 |
float activation_function(float sum) { return sum ; // linear transfer function f(sum)=sum } |
If we sum up all, our simple neuron in a class form can be as below,
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 |
class Tneuron // neuron class { public: float a; // activity of each neurons float w[NN+1]; // weight of links between each neurons Tneuron() { a=0; for(int i=0; i<=NN; i++) w[i]=-1; // if weight is negative there is no link } // let's define a transfer function (or threshold) for the output neuron float activation_function(float sum) { return sum ; // linear transfer function f(sum)=sum } }; |
Now we have a neuron class and each neuron has activity value, weights if it has inputs its own Tneuron() constructor and activation_functin(). We can create many ne[] neuron objects by using this class as below;
|
1 2 3 |
Tneuron ne[NN+1]; // neuron objects |
An activation function can be written as here,
We need a function to fire a nn’th neuron as below,
|
1 2 3 4 5 6 7 8 9 10 11 12 13 |
void fire(int nn) { float sum = 0; for ( int j=0; j<=NN; j++ ) { if( ne[j].w[nn]>=0 ) sum += ne[j].a*ne[j].w[nn]; } ne[nn].a = ne[nn].activation_function(sum); } |
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
#include <stdio.h> #define NN 2 // number of neurons class Tneuron // neuron class { public: float a; // activity of each neurons float w[NN+1]; // weight of links between each neurons Tneuron() { a=0; for(int i=0; i<=NN; i++) w[i]=-1; // if weight is negative there is no link } // let's define an activation function (or threshold) for the output neuron float activation_function(float sum) { return sum ; // linear transfer function f(sum)=sum } }; Tneuron ne[NN+1]; // neuron objects void fire(int nn) { float sum = 0; for ( int j=0; j<=NN; j++ ) { if( ne[j].w[nn]>=0 ) sum += ne[j].a*ne[j].w[nn]; } ne[nn].a = ne[nn].activation_function(sum); } int main() { //let's define activity of two input neurons (a0, a1) and one output neuron (a2) ne[0].a = 0.0; ne[1].a = 1.0; ne[2].a = 0; //let's define weights of signals comes from two input neurons to output neuron (0 to 2 and 1 to 2) ne[0].w[2] = 0.3; ne[1].w[2] = 0.2; // Let's fire our artificial neuron activity, output will be fire(2); printf("%10.6f\n", ne[2].a); getchar(); return 0; } |
This is a very simple class example for an artificial neuron. Classes has many features that you can use professionally. We can create our more functional and complex neuron models. To do this you need to know well about Classes, please read our previous posts about classes.
