/// <summary>
/// Creates instance of genetic algorithm
/// </summary>
/// <param name="graph">Graph on which algorithm will operate</param>
/// <param name="problem">Problem for which algorithm will attempt to find solution</param>
/// <param name="geneticOperators">Genetic operators used for breeding new generations</param>
/// <param name="settings">General settings for solution finding process</param>
/// <param name="startingPopulation">Starting population for algorithm</param>
//TODO: Introduce SavedState complex type
public GeneticAlgorithm(Graph graph, IProblem problem, GeneticOperators geneticOperators,
GeneticAlgorithmSettings settings, ICollection <Individual> startingPopulation = null,
uint currentGeneration = 0, Guid?id = null)
: base(graph, problem, id)
{
if (geneticOperators == null)
{
throw new ArgumentNullException(nameof(geneticOperators));
}
if (!geneticOperators.IsValid())
{
throw new ArgumentException("Genetic operators are not valid", nameof(geneticOperators));
}
if (settings == null)
{
throw new ArgumentNullException(nameof(settings));
}
GeneticOperators = geneticOperators;
Settings = settings;
if (startingPopulation == null || startingPopulation.Count == 0 || !startingPopulation.All(IsIndividualValid))
{
GenerateInitialPopulation();
}
else
{
CurrentPopulation = new SortedSet <Individual>();
foreach (var element in startingPopulation)
{
CurrentPopulation.Add(element);
}
}
CurrentGeneration = currentGeneration;
}
public void test4()
{
Console.WriteLine("Running test 4, cos...");
NeuronSettings nSettings = new NeuronSettings();
NeuronLayerSettings layerSettings = new NeuronLayerSettings(nSettings);
layerSettings.mNumNeuronsPerLayerMin = 1;
layerSettings.mNumNeuronsPerLayerMax = 1;
NeuronNetSettings nnSettings = new NeuronNetSettings(layerSettings);
nnSettings.mNumInputs = 1;
nnSettings.mNumOutputs = 1;
nnSettings.mNumLayersMin = 0;
nnSettings.mNumLayersMax = 0;
GeneticAlgorithmSettings gaSettings = new GeneticAlgorithmSettings(nnSettings.mFactory);
gaSettings.mMaxPopulation = 1000;
GeneticAlgorithm ga = new GeneticAlgorithm(gaSettings);
List <double> inputs = new List <double>()
{
0
};
List <double> expected = new List <double>()
{
0
};
int mNumEpochs = 100;
Console.WriteLine("Learning...");
for (int iEpoch = 0; iEpoch < mNumEpochs; iEpoch++)
{
Console.Write(string.Format("Epoch: {0} / {1}\t", iEpoch + 1, mNumEpochs));
for (double x = -1.5; x <= 1.5; x += 0.001)
{
inputs[0] = x;
expected[0] = Math.Sin(x);
learnPopulation(ga, inputs, expected);
}
Console.Write(string.Format("Average: {0}\t", ga.AverageFitness));
Console.Write(string.Format("Best: {0}\t", ga.BestFitness));
Console.Write(string.Format("Worst: {0}\n", ga.WorstFitness));
ga.epoch();
}
Console.Write(ga.SortedPopulation[0].ToString());
}
static void Main(string[] args)
{
var settings = new GeneticAlgorithmSettings(new BasicStockTrading(), BasicStockTrading.Prices.Count);
settings.StopFunction = new BasicStopFunction {
MinFitness = 335
};
settings.Rng = new StrongRandom(new BufferedRadnomProvider((settings.PopulationSize * settings.GenotypeLength + (3 * settings.PopulationSize))));
var algorithm = new GeneticAlgorithm(settings);
algorithm.Run();
}
public static void GeneticAlgorithmTest()
{
// In this test case, the GA is tasked with finding two numbers A and B
// where A to the power of B = 100 and the difference between A and B is
// as great as possible.
var settings = new GeneticAlgorithmSettings()
{
Objective = OptimisationObjective.Minimise,
MaxGenerations = 1000,
WildcardRate = 0.1,
InitialBoost = 5,
};
var solver = new TestGASolver(settings);
var result = solver.Run();
var value = solver.CalculateValue(result);
Core.Print("Result: " + result[0] + " ^ " + result[1] + " = " + value);
}
/// <summary>
/// Creates instance of genetic algorithm
/// </summary>
/// <param name="graph">Graph on which algorithm will operate</param>
/// <param name="problem">Problem for which algorithm will attempt to find solution</param>
/// <param name="geneticOperators">Genetic operators used for breeding new generations</param>
/// <param name="settings">General settings for solution finding process</param>
/// <param name="startingPopulation">Starting population for algorithm</param>
//TODO: Introduce SavedState complex type
public GeneticAlgorithm(Graph graph, IProblem problem, GeneticOperators geneticOperators,
GeneticAlgorithmSettings settings, ICollection<Individual> startingPopulation = null,
uint currentGeneration = 0, Guid? id = null)
: base(graph, problem, id)
{
if (geneticOperators == null)
throw new ArgumentNullException(nameof(geneticOperators));
if (!geneticOperators.IsValid())
throw new ArgumentException("Genetic operators are not valid", nameof(geneticOperators));
if (settings == null)
throw new ArgumentNullException(nameof(settings));
GeneticOperators = geneticOperators;
Settings = settings;
if (startingPopulation == null || startingPopulation.Count == 0 || !startingPopulation.All(IsIndividualValid))
GenerateInitialPopulation();
else
{
CurrentPopulation = new SortedSet<Individual>();
foreach (var element in startingPopulation)
CurrentPopulation.Add(element);
}
CurrentGeneration = currentGeneration;
}
public TestGASolver(GeneticAlgorithmSettings settings) : base(settings)
{
}
public void test1()
{
Console.WriteLine("Running test 1...");
NeuronSettings nSettings = new NeuronSettings();
NeuronLayerSettings layerSettings = new NeuronLayerSettings(nSettings);
layerSettings.mNumNeuronsPerLayerMin = 2;
layerSettings.mNumNeuronsPerLayerMax = 2;
NeuronNetSettings nnSettings = new NeuronNetSettings(layerSettings);
nnSettings.mNumInputs = 1;
nnSettings.mNumOutputs = 1;
nnSettings.mNumLayersMin = 0;
nnSettings.mNumLayersMax = 0;
GeneticAlgorithmSettings gaSettings = new GeneticAlgorithmSettings(nnSettings.mFactory);
gaSettings.mMaxPopulation = 100;
GeneticAlgorithm ga = new GeneticAlgorithm(gaSettings);
List <double> inputs = new List <double>()
{
0
};
List <double> expected = new List <double>()
{
0
};
int mNumEpochs = 100;
Console.WriteLine("Learning...");
for (int iEpoch = 0; iEpoch < mNumEpochs; iEpoch++)
{
Console.Write(string.Format("Epoch: {0} / {1}\t", iEpoch + 1, mNumEpochs));
for (int x = 0; x <= 100; x++)
{
inputs.Clear();
expected.Clear();
inputs.Add((double)x / 100.0);
expected.Add((double)x / 100.0);
learnPopulation(ga, inputs, expected);
}
Console.Write(string.Format("Average: {0}\t", ga.AverageFitness));
Console.Write(string.Format("Best: {0}\t", ga.BestFitness));
Console.Write(string.Format("Worst: {0}\n", ga.WorstFitness));
ga.epoch();
}
Console.WriteLine("Predicting...");
for (int x = 0; x <= 100; x++)
{
inputs[0] = (double)x / 100.0;
var outputs = predict(ga, inputs);
Console.WriteLine(string.Format("Input:\t{0}\t\tOutput:\t{1}", inputs[0], outputs[0]));
}
}
public void test5Thread()
{
Console.WriteLine("Running test 5, min...");
DateTime timeStart = DateTime.Now;
NeuronSettings nSettings = new NeuronSettings();
NeuronLayerSettings layerSettings = new NeuronLayerSettings(nSettings);
layerSettings.mNumNeuronsPerLayerMin = 2;
layerSettings.mNumNeuronsPerLayerMax = 2;
NeuronNetSettings nnSettings = new NeuronNetSettings(layerSettings);
nnSettings.mNumInputs = 2;
nnSettings.mNumOutputs = 1;
nnSettings.mNumLayersMin = 2;
nnSettings.mNumLayersMax = 2;
GeneticAlgorithmSettings gaSettings = new GeneticAlgorithmSettings(nnSettings.mFactory);
gaSettings.mMaxPopulation = 100;
GeneticAlgorithm ga = new GeneticAlgorithm(gaSettings);
List <double> inputs = new List <double>()
{
0, 0
};
List <double> expected = new List <double>()
{
0
};
int iEpoch = 0;
Console.WriteLine("Learning...");
while (!mLearningStopped && iEpoch < 100)
{
Console.Write(string.Format("Epoch: {0}\t", iEpoch++));
for (double x = 0; x <= 1; x += 0.01)
{
for (double y = 0; y <= 1; y += 0.01)
{
inputs[0] = x;
inputs[1] = y;
expected[0] = x * y;
learnPopulation(ga, inputs, expected);
}
}
Console.Write(string.Format("Average: {0:0.00}\t", ga.AverageFitness));
Console.Write(string.Format("Best: {0:0.00}\t", ga.BestFitness));
Console.Write(string.Format("Worst: {0:0.00}\n", ga.WorstFitness));
ga.epoch();
}
DateTime timeEnd = DateTime.Now;
TimeSpan ts = timeEnd - timeStart;
Console.WriteLine("Learning took: " + ts.ToString());
Console.WriteLine("Predicting...");
for (double x = 0; x < 1; x += 0.1)
{
for (double y = 0; y < 1; y += 0.1)
{
inputs[0] = x;
inputs[1] = y;
var outputs = predict(ga, inputs);
Console.WriteLine(string.Format("Input:\t{0}, {1}\t\tOutput:\t{2:0.000}", inputs[0], inputs[1], outputs[0]));
}
}
}
