//handle network simulation
public void SimulateNetwork()
{
//forward prop network
float[] inputs = new float[] { RaycastForCarScript.sensor1Distance, RaycastForCarScript.sensor2Distance, RaycastForCarScript.sensor3Distance };
float[] outputs = brain.ForwardProp(inputs);
//decide reward
float reward = brain.DecideReward(inputs, outputs);
//backprop data
if (reward > 0)
{
while (reward > 0)
{
brain.BackProp(new float[] { Random.Range(-30F, 30F), Random.Range(-30F, 30F), Random.Range(-30F, 30F) }, outputs);
reward--;
}
}
else if (reward < 0)
{
while (reward < 0)
{
brain.BackProp(inputs, new float[] { Random.Range(-15F, 15F), Random.Range(-15F, 15F) });
reward++;
}
}
//add output to memory bank
brain.MemoryBank.AddMemory(inputs, outputs);
//move car
float engine = outputs[0];
float turn = outputs[1];
if (engine > MaxSpeed)
{
engine = MaxSpeed;
}
if (engine < -MaxSpeed)
{
engine = -MaxSpeed;
}
if (turn > MaxTurn)
{
turn = MaxTurn;
}
if (turn < -MaxTurn)
{
turn = -MaxTurn;
}
MoveCar(engine, turn);
}
private void Learning(RichTextBox rtb, ProgressBar pb, Button btn_start)
{
this.Dispatcher.Invoke(() => rtb.AppendText(String.Format("Training...(0/{0})\n\n", this.reps)));
for (int j = 0; j < this.reps; j++)
{
for (int i = 0; i < inputs.Count; i++)
{
try
{
lock (nn)
{
nn.FeedForward(inputs[i]);
nn.BackProp(outputs[i]);
}
}
catch (Exception)
{
this.Dispatcher.Invoke(() => rtb.AppendText("Failed"));
}
}
this.Dispatcher.Invoke(() => rtb.AppendText(String.Format("Training...({0}/{1})\n", j + 1, this.reps)));
this.Dispatcher.Invoke(() => rtb_log.ScrollToEnd());
this.Dispatcher.Invoke(() => pb.Value = j + 1);
}
this.Dispatcher.Invoke(() => btn_start.IsEnabled = true);
this.Dispatcher.Invoke(() => pb.Value = 0);
this.Dispatcher.Invoke(() => rtb_log.AppendText("Completed!\n"));
t1 = new Thread(() => Learning(this.rtb_log, this.pb_training, this.btn_start));
}
public void Calcular()
{
float[][] entrada;
entrada = new float[TamañoBatch][];
float[][] esperado;
esperado = new float[TamañoBatch][];
for (int i = 0; i < entrada.Length; i++)
{
entrada[i] = new float[28 * 28];
esperado[i] = new float[10];
}
for (int i = 0; i < 60000 / TamañoBatch; i++) // 60 000 -> 59 900
{
for (int j = 0; j < TamañoBatch; j++) // pasa la informacion de entrada
{
entrada[j] = mnist.valoresDeEntrenamoento[(i * TamañoBatch) + j];
float[] lab = new float[10];
lab[(int)mnist.lablesEntrenamiento[(i * TamañoBatch) + j]] = 1;
esperado[j] = lab;
net.FeedForward(entrada[j]);
net.BackProp(esperado[j]);
net.UpdateNetwork();
}
//net.UpdateNetwork();
}
Probar();
}
// Use this for initialization
void Start()
{
// 0 0 0 => 0
// 0 0 1 => 1
// 0 1 0 => 1
// 0 1 1 => 0
// 1 0 0 => 1
// 1 0 1 => 0
// 1 1 0 => 0
// 1 1 1 => 1
NeuralNetwork net = new NeuralNetwork(new int[] { 3, 25, 25, 1 }); //intiilize network
//Itterate 5000 times and train each possible output
//5000*8 = 40000 traning operations
for (int i = 0; i < 5000; i++)
{
net.FeedForward(new float[] { 0, 0, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 0, 0, 1 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, 1, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, 1, 1 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 0, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 1, 0, 1 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 1, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 1, 1 });
net.BackProp(new float[] { 1 });
}
//output to see if the network has learnt
//WHICH IT HAS!!!!!
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 0, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 0, 1 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 1, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 1, 1 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 1, 0, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 1, 0, 1 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 1, 1, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 1, 1, 1 })[0]);
}
static void Main(string[] args)
{
Console.WriteLine("Welcome!\nour Neural Network is now to Study the 3-Input XOR Pattern!");
Console.WriteLine("Welcome!\nwe first start by training the Neural Network for you!");
NeuralNetwork nn = new NeuralNetwork(new int[] { 3, 25, 25, 1 });
//Neural Network Training!
for (int i = 0; i < epochs; i++)
{
nn.FeedForward(new float[] { 0, 0, 0 }); //INPUT LAYER of 3 Neurons
nn.BackProp(new float[] { 0 }); //HAS THIS Expected Output Layer of 1 Neuron
nn.FeedForward(new float[] { 0, 0, 1 });
nn.BackProp(new float[] { 1 });
nn.FeedForward(new float[] { 0, 1, 0 });
nn.BackProp(new float[] { 1 });
nn.FeedForward(new float[] { 0, 1, 1 });
nn.BackProp(new float[] { 0 });
nn.FeedForward(new float[] { 1, 0, 0 });
nn.BackProp(new float[] { 1 });
nn.FeedForward(new float[] { 1, 0, 1 });
nn.BackProp(new float[] { 0 });
nn.FeedForward(new float[] { 1, 1, 0 });
nn.BackProp(new float[] { 0 });
nn.FeedForward(new float[] { 1, 1, 1 });
nn.BackProp(new float[] { 1 });
}
//Finished Training
Console.WriteLine("Finished Training!!!");
Console.WriteLine("Testing Area:\n");
var o = nn.FeedForward(new float[] { 0, 0, 0 });
Console.WriteLine("the Output for Sequence 0 0 0 is: " + o[0]);
o = nn.FeedForward(new float[] { 0, 0, 1 });
Console.WriteLine("the Output for Sequence 0 0 1 is: " + o[0]);
o = nn.FeedForward(new float[] { 1, 0, 1 });
Console.WriteLine("the Output for Sequence 1 0 1 is: " + o[0]);
o = nn.FeedForward(new float[] { 1, 1, 1 });
Console.WriteLine("the Output for Sequence 1 1 1 is: " + o[0]);
Console.WriteLine("Done! Press Any Key to Close...");
Console.ReadKey();
}
void Start()
{
// 0 0 0 => 0
// 0 0 1 => 1
// 0 1 0 => 1
// 0 1 1 => 0
// 1 0 0 => 1
// 1 0 1 => 0
// 1 1 0 => 0
// 1 1 1 => 1
NeuralNetwork net = new NeuralNetwork(new int[] { 3, 25, 25, 1 });
float n = 10f;
for (int i = 0; i < 5000; i++)
{
net.FeedForward(new float[] { 0, 0, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 0, 0, n });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, n, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, n, n });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { n, 0, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { n, 0, n });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { n, n, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { n, n, n });
net.BackProp(new float[] { 1 });
}
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 0, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, 0, n })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, n, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { 0, n, n })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { n, 0, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { n, 0, n })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { n, n, 0 })[0]);
UnityEngine.Debug.Log(net.FeedForward(new float[] { n, n, n })[0]);
}
public static void Learning(NeuralNetwork net, string fileName)
{
try
{
LoadInputOutput(fileName);
}
catch { throw new Exception("Nem lehetett beolvasni a fájlból"); }
for (int i = 0; i < inputs.Count; i++)
{
net.FeedForward(inputs[i]);
net.BackProp(outputs[i]);
}
}
private void OnCollisionEnter(Collision collision)
{
corrector = Vector3.Reflect(corrector, collision.contacts[0].normal);
network.BackProp(rb.velocity, corrector);
transform.position = pos;
}
public void TrainNetwork() //When user presses train button, this function will be run and network will be trained and saved in file.
{
// Read in training values:
string path = "Assets/Resources/TrainingValues.txt";
// TestValues:
float[][] input = new float[256][];
float[][] output = new float[256][];
//Read the text from directly from the test.txt file
StreamReader reader = new StreamReader(path);
string line;
int inputNr = 0;
int outputNr = 0;
int lineNr = 0;
while ((line = reader.ReadLine()) != null)
{
if (line.StartsWith("#")) // Don't read comments in the txt file:
{
continue;
}
if (lineNr % 2 == 0)
{
string[] bits = line.Split(' ');
float[] array = new float[bits.Length];
for (int i = 0; i < bits.Length; i++)
{
array[i] = float.Parse(bits[i]);
}
input[inputNr] = array;
inputNr++;
}
else
{
string[] bits = line.Split(' ');
float[] array = new float[bits.Length];
for (int i = 0; i < bits.Length; i++)
{
array[i] = float.Parse(bits[i]);
}
output[outputNr] = array;
outputNr++;
}
lineNr++;
}
NeuralNetwork net = new NeuralNetwork(new int[] { 10, 25, 25, 4 }); // initialize network
for (int i = 0; i < trainGenerations; i++)
{
for (int j = 0; j < (lineNr / 2); j++)
{
net.FeedForward(input[j]);
net.BackProp(output[j]);
}
}
// Used to see the accuracy of the network:
float accuracy = 0;
for (int i = 0; i < trainGenerations; i++)
{
for (int j = 0; j < (lineNr / 2); j++)
{
float[] outputTest = net.FeedForward(input[j]);
for (int k = 0; k < 4; k++)
{
if (output[j][k] == 1) // Supposed to be 1.
{
float newValue = outputTest[k] * 2 - 1; // Making value from 0.5-1 to 0-1.
accuracy = (accuracy + newValue) / 2;
}
else // Supposed to be 0.
{
float newValue = 1 - (outputTest[k] * 2); // Making value from 0.5-0 to 0-1.
accuracy = (accuracy + newValue) / 2;
}
}
}
}
Debug.Log("Accuracy: " + accuracy);
// Save the training values:
net.SaveBrain();
net.LoadBrain();
}
static void Main(string[] args)
{
// 0 0 0 => 0
// 0 0 1 => 1
// 0 1 0 => 1
// 0 1 1 => 0
// 1 0 0 => 1
// 1 0 1 => 0
// 1 1 0 => 0
// 1 1 1 => 1
NeuralNetwork net = new NeuralNetwork(new int[] { 3, 25, 25, 1 });
//Itterate 5000 times and train each possible output
//5000*8 = 40000 traning operations
for (int i = 0; i < 5000; i++)
{
net.FeedForward(new float[] { 0, 0, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 0, 0, 1 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, 1, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 0, 1, 1 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 0, 0 });
net.BackProp(new float[] { 1 });
net.FeedForward(new float[] { 1, 0, 1 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 1, 0 });
net.BackProp(new float[] { 0 });
net.FeedForward(new float[] { 1, 1, 1 });
net.BackProp(new float[] { 1 });
}
Console.WriteLine(net.FeedForward(new float[] { 0, 0, 0 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 0, 0, 1 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 0, 1, 0 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 0, 1, 1 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 1, 0, 0 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 1, 0, 1 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 1, 1, 0 })[0]);
Console.WriteLine(net.FeedForward(new float[] { 1, 1, 1 })[0]);
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 0, 0, 0 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 0, 0, 1 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 0, 1, 0 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 0, 1, 1 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 1, 0, 0 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 1, 0, 1 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 1, 1, 0 })[0]));
//Console.WriteLine(Math.Round(net.FeedForward(new float[] { 1, 1, 1 })[0]));
Console.WriteLine("\nPress Enter to quit.");
Console.ReadLine();
}
