484 lines
12 KiB
C++
484 lines
12 KiB
C++
#include <iostream>
|
||
#include <cstdlib>
|
||
#include <time.h>
|
||
|
||
|
||
float RandomFloat(int min, int max)
|
||
{
|
||
static unsigned long int counter = 0;
|
||
srand(time(0) + counter++ * 50);
|
||
int value = (rand() % ((max - min) * 100));
|
||
return (float)value / 100.0 + (float)min + 1.0;
|
||
}
|
||
|
||
#pragma region Sinaps
|
||
class Sinaps
|
||
{
|
||
private:
|
||
float weight; // Ağırlık
|
||
float value; // Değer
|
||
float bias; // Öteleme
|
||
public:
|
||
Sinaps();
|
||
Sinaps(float, float, float); // Kaydedilen değerleri yeniden yazabilmek için
|
||
void SetSinaps(float, float, float); // Sonradan tamamen değiştirebilmek için
|
||
void SetWeight(float);
|
||
void SetValue(float);
|
||
void SetBias(float);
|
||
float Fire();
|
||
};
|
||
|
||
Sinaps::Sinaps() { weight = value = bias = 0.0; }
|
||
Sinaps::Sinaps(float weight, float value, float bias)
|
||
{
|
||
this -> weight = weight;
|
||
this -> value = value;
|
||
this -> bias = bias;
|
||
}
|
||
|
||
void Sinaps::SetSinaps(float weight, float value, float bias) {
|
||
std::cout << "weight = " << weight << "\n";
|
||
std::cout << "value = " << value << "\n";
|
||
std::cout << "bias = " << bias << "\n";
|
||
Sinaps(weight, value, bias); }
|
||
void Sinaps::SetWeight(float weight) { this -> weight = weight; }
|
||
void Sinaps::SetValue(float value) { std::cout << value << "\n"; this -> value = value; }
|
||
void Sinaps::SetBias(float bias) { this -> bias = bias; }
|
||
|
||
float Sinaps::Fire() { return weight * value + bias; }
|
||
#pragma endregion
|
||
#pragma region Noron
|
||
class Noron
|
||
{
|
||
private:
|
||
Sinaps *forwards;
|
||
Sinaps *incoming;
|
||
int forwardsCount;
|
||
int incomingCount;
|
||
public:
|
||
Noron();
|
||
~Noron();
|
||
bool SetForwards(Sinaps *, int);
|
||
bool SetIncoming(Sinaps *, int);
|
||
float GetStatus();
|
||
};
|
||
|
||
Noron::Noron()
|
||
{
|
||
forwards = incoming = NULL;
|
||
forwardsCount = incomingCount = 0;
|
||
}
|
||
|
||
Noron::~Noron()
|
||
{
|
||
delete forwards;
|
||
delete incoming;
|
||
}
|
||
|
||
bool Noron::SetForwards(Sinaps *newForwards, int size)
|
||
{
|
||
forwards = (Sinaps *) new char[sizeof(Sinaps) * size];
|
||
|
||
if(!forwards) return false;
|
||
|
||
for (int i = 0; i < size; i++)
|
||
*(forwards+i) = *(newForwards+i);
|
||
|
||
forwardsCount = size;
|
||
return true;
|
||
}
|
||
|
||
bool Noron::SetIncoming(Sinaps *newIncoming, int size)
|
||
{
|
||
incoming = (Sinaps *) new char[sizeof(Sinaps) * size];
|
||
|
||
if(!incoming) return false;
|
||
|
||
for (int i = 0; i < size; i++)
|
||
*(incoming+i) = *(newIncoming+i);
|
||
|
||
incomingCount = size;
|
||
return true;
|
||
}
|
||
|
||
float Noron::GetStatus()
|
||
{
|
||
float toplam = 0.0;
|
||
std::cout << toplam << "\n";
|
||
for (int i = 0; i < incomingCount; i++)
|
||
toplam += (incoming + i) -> Fire();
|
||
|
||
for (int i = 0; i < forwardsCount; i++)
|
||
(forwards + i) -> SetValue(toplam);
|
||
|
||
return toplam;
|
||
}
|
||
#pragma endregion
|
||
#pragma region Katman
|
||
class Katman
|
||
{
|
||
protected:
|
||
Noron *neurons;
|
||
Katman *forward;
|
||
Sinaps *layerSinapses;
|
||
int size;
|
||
Sinaps *CreateSinapsSet(int size);
|
||
public:
|
||
Katman();
|
||
Katman(int);
|
||
~Katman();
|
||
void FireLayer();
|
||
void RandomizeSinapsValues();
|
||
bool SetForward(Katman *);
|
||
bool SetIncoming(Sinaps *sinapsSet, int backwardsNeuronCount);
|
||
bool SetNoron(Noron *, int);
|
||
bool CreateNoron(int);
|
||
int GetSize();
|
||
};
|
||
|
||
Katman::Katman() { neurons = NULL; this -> size = 0; }
|
||
Katman::Katman(int size)
|
||
{
|
||
Katman();
|
||
|
||
if(!CreateNoron(size))
|
||
std::cout << "Katman Oluşturulamadı!";
|
||
else
|
||
this -> size = size;
|
||
}
|
||
Katman::~Katman() { delete neurons; }
|
||
|
||
|
||
Sinaps *Katman::CreateSinapsSet(int size)
|
||
{
|
||
Sinaps* sinapses = (Sinaps *) new char[sizeof(Sinaps) * size];
|
||
|
||
if(sinapses)
|
||
for (int i = 0; i < size; i++)
|
||
*(sinapses + i) = Sinaps();
|
||
|
||
return sinapses;
|
||
}
|
||
void Katman::RandomizeSinapsValues()
|
||
{
|
||
if(!forward)
|
||
return;
|
||
|
||
int sinapsCount = size * (forward -> GetSize());
|
||
for (int i = 0; i < sinapsCount; i++)
|
||
{
|
||
(layerSinapses + i) -> SetSinaps(
|
||
RandomFloat(-5, 5),
|
||
RandomFloat(-5, 5),
|
||
RandomFloat(-5, 5)
|
||
);
|
||
}
|
||
}
|
||
|
||
void Katman::FireLayer()
|
||
{
|
||
for (int i = 0; i < size; i++)
|
||
std::cout << i << ". Fire = " << (neurons + i) -> GetStatus() << "\n";
|
||
}
|
||
|
||
bool Katman::SetForward(Katman *forward)
|
||
{
|
||
Sinaps *sinapses = NULL; // Pointer to store all created sinapses
|
||
Sinaps *newSinapses = NULL; // Temporary Pointer for creating each neurons' s1inapses
|
||
int forwardSize;
|
||
int sinapsesIndex = 0;
|
||
delete layerSinapses;
|
||
|
||
this -> forward = forward;
|
||
|
||
if(!forward)
|
||
return true;
|
||
|
||
forwardSize = forward -> GetSize();
|
||
|
||
sinapses = (Sinaps *) new char[sizeof(Sinaps) * size * forwardSize];
|
||
|
||
if(!sinapses)
|
||
return false;
|
||
|
||
// Set Forwards of each neuron in the Layer
|
||
for (int thisCounter = 0; thisCounter < size; thisCounter++)
|
||
{
|
||
newSinapses = CreateSinapsSet(forwardSize);
|
||
|
||
if(!newSinapses)
|
||
return false;
|
||
|
||
(neurons + thisCounter) -> SetForwards(newSinapses, forwardSize);
|
||
|
||
// Add each sinaps to the array
|
||
for (int forwardCounter = 0; forwardCounter < forwardSize; forwardCounter++)
|
||
*(sinapses + (sinapsesIndex++)) = *(newSinapses + forwardCounter);
|
||
}
|
||
|
||
layerSinapses = sinapses;
|
||
// Send the sinapses to the forward layer
|
||
return forward -> SetIncoming(sinapses, size);
|
||
}
|
||
|
||
bool Katman::SetIncoming(Sinaps *sinapsSet, int backwardsNeuronCount)
|
||
{
|
||
Sinaps *sinapses = NULL;
|
||
|
||
sinapses = (Sinaps *) new char[sizeof(Sinaps) * backwardsNeuronCount];
|
||
|
||
if(!sinapses)
|
||
return false;
|
||
|
||
for (int thisCounter = 0; thisCounter < size; thisCounter++)
|
||
{
|
||
// Add each sinaps to the array
|
||
for (int incomingCounter = 0; incomingCounter < backwardsNeuronCount; incomingCounter++)
|
||
*(sinapses + (size * thisCounter + incomingCounter)) = *(sinapsSet + incomingCounter);
|
||
|
||
(neurons + thisCounter) -> SetIncoming(sinapses, backwardsNeuronCount);
|
||
}
|
||
|
||
return true;
|
||
}
|
||
|
||
bool Katman::SetNoron(Noron *newneurons, int size)
|
||
{
|
||
neurons = (Noron *) new char[sizeof(Noron) * size];
|
||
|
||
if(!neurons) return false;
|
||
|
||
for (int i = 0; i < size; i++)
|
||
*(neurons+i) = *(newneurons+i);
|
||
|
||
this -> size = size;
|
||
return true;
|
||
}
|
||
|
||
bool Katman::CreateNoron(int size)
|
||
{
|
||
neurons = (Noron *) new char[sizeof(Noron) * size];
|
||
|
||
if(!neurons) return false;
|
||
|
||
for (int i = 0; i < size; i++)
|
||
*(neurons+i) = Noron();
|
||
|
||
this -> size = size;
|
||
return true;
|
||
}
|
||
|
||
int Katman::GetSize() { return size; }
|
||
#pragma endregion
|
||
#pragma region Girdi-Cikti
|
||
#pragma region Girdi
|
||
class Girdi : public Katman
|
||
{
|
||
public:
|
||
Girdi();
|
||
Girdi(int);
|
||
void SetValue(int, float);
|
||
};
|
||
|
||
Girdi::Girdi() : Katman() {}
|
||
Girdi::Girdi(int size) : Katman(size) {}
|
||
|
||
void Girdi::SetValue(int index, float value)
|
||
{
|
||
Sinaps *editedSinaps = NULL;
|
||
int forwardNeuronCount = forward -> GetSize();
|
||
|
||
for (int i = 0; i < forwardNeuronCount; i++)
|
||
{
|
||
editedSinaps = (layerSinapses + index * forwardNeuronCount + i);
|
||
editedSinaps -> SetValue(value);
|
||
}
|
||
}
|
||
#pragma endregion
|
||
#pragma region Cikti
|
||
class Cikti : public Katman
|
||
{
|
||
public:
|
||
Cikti();
|
||
Cikti(int);
|
||
float GetValue(int);
|
||
};
|
||
|
||
Cikti::Cikti() : Katman() {}
|
||
Cikti::Cikti(int size) : Katman(size) {}
|
||
|
||
float Cikti::GetValue(int index)
|
||
{
|
||
return (neurons + index) -> GetStatus();
|
||
}
|
||
#pragma endregion
|
||
#pragma endregion
|
||
#pragma region NeuralNetwork
|
||
class NeuralNetwork
|
||
{
|
||
private:
|
||
Girdi *input;
|
||
Katman *hiddenLayers;
|
||
Cikti *output;
|
||
int hiddenSize;
|
||
public:
|
||
NeuralNetwork();
|
||
NeuralNetwork(int);
|
||
~NeuralNetwork();
|
||
void SetInput(int, float);
|
||
void RandomizeNetworkValues();
|
||
void FireNetwork();
|
||
bool SetInputNeurons(int);
|
||
bool SetHiddenLayerNeurons(int, int);
|
||
bool SetOutputNeurons(int);
|
||
bool ConnectLayers();
|
||
float GetOutputValue(int);
|
||
};
|
||
|
||
NeuralNetwork::NeuralNetwork()
|
||
{
|
||
hiddenSize = 0;
|
||
input = NULL;
|
||
hiddenLayers = NULL;
|
||
output = NULL;
|
||
}
|
||
|
||
NeuralNetwork::NeuralNetwork(int hiddenSize)
|
||
{
|
||
input = new Girdi();
|
||
hiddenLayers = new Katman[hiddenSize];
|
||
output = new Cikti();
|
||
|
||
if(!input)
|
||
std::cout << "Girdi Katmani Olusturulamadi!" << "\n";
|
||
if(!hiddenLayers)
|
||
std::cout << "Ara Katmanlar Olusturulamadi!" << "\n";
|
||
if(!output)
|
||
std::cout << "Cikti Katmani Olusturulamadi!" << "\n";
|
||
|
||
if(!input || !hiddenLayers || !output)
|
||
return;
|
||
|
||
this -> hiddenSize = hiddenSize;
|
||
}
|
||
|
||
NeuralNetwork::~NeuralNetwork()
|
||
{
|
||
delete input;
|
||
delete hiddenLayers;
|
||
delete output;
|
||
}
|
||
|
||
bool NeuralNetwork::SetHiddenLayerNeurons(int index, int size)
|
||
{
|
||
Noron *neurons = new Noron[size];
|
||
|
||
if(!neurons)
|
||
return false;
|
||
|
||
return (hiddenLayers + index) -> SetNoron(neurons, size);
|
||
}
|
||
|
||
bool NeuralNetwork::SetInputNeurons(int size)
|
||
{
|
||
Noron *neurons = new Noron[size];
|
||
|
||
if(!neurons)
|
||
return false;
|
||
|
||
return output -> SetNoron(neurons, size);
|
||
}
|
||
|
||
bool NeuralNetwork::SetOutputNeurons(int size)
|
||
{
|
||
Noron *neurons = new Noron[size];
|
||
|
||
if(!neurons)
|
||
return false;
|
||
|
||
return input -> SetNoron(neurons, size);
|
||
}
|
||
|
||
bool NeuralNetwork::ConnectLayers()
|
||
{
|
||
if(!input -> SetForward(hiddenLayers))
|
||
return false;
|
||
|
||
for (int i = 0; i < hiddenSize - 1; i++)
|
||
if(!(hiddenLayers + i) -> SetForward(hiddenLayers + i + 1))
|
||
return false;
|
||
|
||
if(!(hiddenLayers + hiddenSize - 1) -> SetForward(output))
|
||
return false;
|
||
|
||
return output -> SetForward(NULL);
|
||
}
|
||
|
||
void NeuralNetwork::FireNetwork()
|
||
{
|
||
input -> FireLayer();
|
||
|
||
for (int i = 0; i < hiddenSize; i++)
|
||
(hiddenLayers + i) -> FireLayer();
|
||
|
||
output -> FireLayer();
|
||
}
|
||
|
||
void NeuralNetwork::SetInput(int index, float value)
|
||
{
|
||
if(!input)
|
||
return;
|
||
|
||
input -> SetValue(index, value);
|
||
}
|
||
|
||
void NeuralNetwork::RandomizeNetworkValues()
|
||
{
|
||
input -> RandomizeSinapsValues();
|
||
|
||
for (int i = 0; i < hiddenSize; i++)
|
||
(hiddenLayers + i) -> RandomizeSinapsValues();
|
||
|
||
output -> RandomizeSinapsValues();
|
||
}
|
||
|
||
float NeuralNetwork::GetOutputValue(int index)
|
||
{
|
||
return output -> GetValue(index);
|
||
}
|
||
#pragma endregion
|
||
int main(int argc, char const *argv[])
|
||
{
|
||
NeuralNetwork network(3);
|
||
|
||
network.SetInputNeurons(1);
|
||
network.SetHiddenLayerNeurons(0, 2);
|
||
network.SetHiddenLayerNeurons(1, 3);
|
||
network.SetHiddenLayerNeurons(2, 2);
|
||
network.SetOutputNeurons(1);
|
||
|
||
network.ConnectLayers();
|
||
network.RandomizeNetworkValues();
|
||
|
||
network.SetInput(0, 1);
|
||
std::cout << "m1\n";
|
||
|
||
network.FireNetwork();
|
||
std::cout << "m2\n";
|
||
std::cout << network.GetOutputValue(0) << "\n";
|
||
std::cout << "m3\n";
|
||
|
||
|
||
|
||
|
||
|
||
// Katman k1(5);
|
||
// Katman k2(3);
|
||
|
||
// std::cout << "k1 SetForward = " << k1.SetForward(&k2) << "\n";
|
||
// std::cout << "k2 SetForward = " << k2.SetForward(NULL) << "\n";
|
||
// k1.FireLayer();
|
||
// k2.FireLayer();
|
||
return 0;
|
||
}
|