public void TestActivate()
{
perceptron.Feed(Input.FromArray(new float[] { 1, 1, 1, 1 }));
float expect = 1;
float actual = perceptron.Activate();
Assert.AreEqual(expect, actual);
}
static void Evaluate(Perceptron p, double from, double to, double step)
{
string output = "";
for (double i = from; i < to; i += step)
{
double res = p.Activate(new double[] { normalize(i, inputMin, inputMax) })[0];
output += i + ";" + inverseNormalize(res, outputMin, outputMax) + "\n";
Console.WriteLine(i + ";" + res + "\n");
}
System.IO.File.WriteAllText(outputPath, output);
}
//Tests an Individual's "solution" (the weights and bias it generated)
static int RunTest(ref Perceptron perceptron, DataSet dataSet, Individual individual)
{
int score = 0;
ApplyWeightAndBias(ref perceptron, individual);
for (int i = 0; i < dataSet.Data.Count; i++)
{
double perceptronResult =
Math.Round(perceptron.Activate(dataSet.Data[i]), MidpointRounding.AwayFromZero);
if (perceptronResult == dataSet.DataSolutions[i])
{
score++;
}
}
return(score);
}
static void Main(string[] args)
{
//Start prompt.
Console.Title = "Evolutionary Perceptron";
Console.WriteLine("Evolutionary Perceptron is ready. Press enter to start...");
Console.ReadLine();
//Initialize everything.
Random random = new Random();
DataSet dataSet = new DataSet("Training data\\CloudsDataSet.txt");
GeneticAlgorithm geneticAlgorithm = new GeneticAlgorithm(500, 3, 0.5d, random);
Perceptron perceptron = new Perceptron(2);
//Load info about the data set.
Console.WriteLine("Loaded data set:\n" + dataSet.Name + "\n");
Console.WriteLine("Description:\n" + dataSet.Description);
//Evolve the perceptron.
while (true)
{
/* Test all the Individuals (which evolve the weights and bias
* for the perceptron) to see which ones have generated the best
* weights and bias for the perceptron. */
for (int i = 0; i < geneticAlgorithm.Population.Length; i++)
{
geneticAlgorithm.Population[i].Fitness =
RunTest(ref perceptron, dataSet, geneticAlgorithm.Population[i]);
}
Console.Write("\rGeneration: " + geneticAlgorithm.Generation +
", Fitness: " + geneticAlgorithm.BestIndividual.Fitness);
//Check to exit the loop (see note at the bottom of this file).
if (geneticAlgorithm.BestIndividual.Fitness >= dataSet.Data.Count)
{
ApplyWeightAndBias(ref perceptron, geneticAlgorithm.BestIndividual);
break;
}
geneticAlgorithm.EvolvePopulation();
}
Console.WriteLine("\n\nRunning test data...\n");
//Run the test data through the perceptron.
foreach (double[] testData in dataSet.TestData)
{
double perceptronResult = perceptron.Activate(testData);
string cloudShape = testData[1] == 0 ? "puffy" : "thin";
string cloudAltitude = testData[0].ToString();
string cloudType = perceptronResult < 0.5 ? "Cumulus" : "Cirrus";
Console.WriteLine("Test data: " + cloudAltitude + "," + cloudShape);
Console.WriteLine("A cloud that is " + cloudShape +
" in shape and has an altitude of " + cloudAltitude +
" is a " + cloudType + " cloud.\n");
}
Console.WriteLine("\nExecution finished. Press enter to exit...");
Console.ReadLine();
}
