static void Main(string[] args)
{
// this form of ANN uses genetic algorithm to produce
// hidden layer of neurons
// A NEAT network starts with only an
// input layer and output layer. The rest is evolved as the training progresses.
// Connections inside of a NEAT neural network can be feedforward, recurrent,
// or self - connected.All of these connection types will be tried by NEAT as it
// attempts to evolve a neural network capable of the given task.
IMLDataSet trainingSet = new BasicMLDataSet(XORInput, XORIdeal);
NEATPopulation pop = new NEATPopulation(2, 1, 1000);
pop.Reset();
pop.InitialConnectionDensity = 1.0; // not required, but speeds processing.
ICalculateScore score = new TrainingSetScore(trainingSet);
// train the neural network
TrainEA train = NEATUtil.ConstructNEATTrainer(pop, score);
EncogUtility.TrainToError(train, 0.01);
NEATNetwork network = (NEATNetwork)train.CODEC.Decode(train.BestGenome);
// TODO no persistance? no means to peek structure?
// test the neural network
Console.WriteLine(@"Neural Network Results:");
EncogUtility.Evaluate(network, trainingSet);
}
public void Eval(String start, String expect)
{
EncogProgramContext context = new EncogProgramContext();
StandardExtensions.CreateNumericOperators(context);
PrgPopulation pop = new PrgPopulation(context, 1);
ICalculateScore score = new ZeroEvalScoreFunction();
TrainEA genetic = new TrainEA(pop, score);
genetic.ValidationMode = true;
genetic.CODEC = new PrgCODEC();
genetic.AddOperation(0.95, new SubtreeCrossover());
genetic.AddOperation(0.05, new SubtreeMutation(context, 4));
genetic.AddScoreAdjuster(new ComplexityAdjustedScore());
genetic.Rules.AddRewriteRule(new RewriteConstants());
genetic.Rules.AddRewriteRule(new RewriteAlgebraic());
EncogProgram expression = new EncogProgram(context);
expression.CompileExpression(start);
RenderCommonExpression render = new RenderCommonExpression();
genetic.Rules.Rewrite(expression);
Assert.AreEqual(expect, render.Render(expression));
}
public TrainEA RunGA(TrainEA geneticAlgorithm, int maxSameSolutionCount)
{
try
{
var sameSolutionCount = 0;
var lastSolutionError = double.MaxValue;
var iteration = 1;
while (sameSolutionCount < maxSameSolutionCount)
{
geneticAlgorithm.Iteration();
var currentSolutionError = geneticAlgorithm.Error;
if (Math.Abs(lastSolutionError - currentSolutionError) < 1.0)
{
sameSolutionCount++;
}
else
{
sameSolutionCount = 0;
}
lastSolutionError = currentSolutionError;
Console.WriteLine($"{iteration++} Iteration - Error {currentSolutionError}");
}
geneticAlgorithm.FinishTraining();
Console.WriteLine("Good Solutions Found");
return(geneticAlgorithm);
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
/// <summary>
/// Create an NEAT GA trainer.
/// </summary>
/// <param name="method">The method to use.</param>
/// <param name="training">The training data to use.</param>
/// <param name="argsStr">The arguments to use.</param>
/// <returns>The newly created trainer.</returns>
public IMLTrain Create(IMLMethod method,
IMLDataSet training, String argsStr)
{
ICalculateScore score = new TrainingSetScore(training);
TrainEA train = NEATUtil.ConstructNEATTrainer((NEATPopulation)method, score);
return(train);
}
public int[] Run()
{
StringBuilder builder = new StringBuilder();
IPopulation pop = InitPopulation();
ICalculateScore score = new FitnessFunction(cities);
genetic = new TrainEA(pop, score);
genetic.AddOperation(0.9, new SpliceNoRepeat(cityCount / 3));
genetic.AddOperation(0.1, new MutateShuffle());
int sameSolutionCount = 0;
int iteration = 1;
double lastSolution = Double.MaxValue;
while (sameSolutionCount < maxSameSolition)
{
genetic.Iteration();
double thisSolution = genetic.Error;
builder.Length = 0;
builder.Append("Iteration: ");
builder.Append(iteration++);
builder.Append(", Best Path Length = ");
builder.Append(thisSolution);
Console.WriteLine(builder.ToString());
if (Math.Abs(lastSolution - thisSolution) < 1.0)
{
sameSolutionCount++;
}
else
{
sameSolutionCount = 0;
}
lastSolution = thisSolution;
}
Console.WriteLine("Good solution found:");
IntegerArrayGenome best = (IntegerArrayGenome)genetic.BestGenome;
genetic.FinishTraining();
return(best.Data);
}
/// <summary>
/// Setup and solve the TSP.
/// </summary>
public void Execute(IExampleInterface app)
{
this.app = app;
var builder = new StringBuilder();
initCities();
IPopulation pop = initPopulation();
ICalculateScore score = new TSPScore(cities);
genetic = new TrainEA(pop, score);
genetic.AddOperation(0.9, new SpliceNoRepeat(CITIES / 3));
genetic.AddOperation(0.1, new MutateShuffle());
int sameSolutionCount = 0;
int iteration = 1;
double lastSolution = Double.MaxValue;
while (sameSolutionCount < MAX_SAME_SOLUTION)
{
genetic.Iteration();
double thisSolution = genetic.Error;
builder.Length = 0;
builder.Append("Iteration: ");
builder.Append(iteration++);
builder.Append(", Best Path Length = ");
builder.Append(thisSolution);
Console.WriteLine(builder.ToString());
if (Math.Abs(lastSolution - thisSolution) < 1.0)
{
sameSolutionCount++;
}
else
{
sameSolutionCount = 0;
}
lastSolution = thisSolution;
}
Console.WriteLine("Good solution found:");
displaySolution();
genetic.FinishTraining();
}
public TrainEA CreateGA(int cityCount, BasicPopulation population, TravellingSalesManFitnessScore tspFitness, double crossOverProbabality, double mutationProbabality)
{
try
{
var geneticAlgorithm = new TrainEA(population, tspFitness);
geneticAlgorithm.AddOperation(crossOverProbabality, new SpliceNoRepeat(cityCount / 3));
geneticAlgorithm.AddOperation(mutationProbabality, new MutateShuffle());
return(geneticAlgorithm);
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
public TrainEA CreateGA(int powerUnitCount, BasicPopulation population, PowerUnitMaintainanceFitnessFunction maintainanceFitness, double crossOverProbabality, double mutationProbabality)
{
try
{
var geneticAlgorithm = new TrainEA(population, maintainanceFitness);
geneticAlgorithm.AddOperation(crossOverProbabality, new CustomCrossOver());
geneticAlgorithm.AddOperation(mutationProbabality, new CustomMutation());
return(geneticAlgorithm);
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
public void DisplaySolution(TrainEA geneticAlgorithm, City [] cities)
{
try
{
var bestChromosome = (IntegerArrayGenome)geneticAlgorithm.Population.BestGenome;
foreach (var gene in bestChromosome.Data)
{
Console.WriteLine($"city - {gene} => X = {cities[gene].X}, Y = {cities[gene].Y}");
}
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
/// <summary>
/// Create an EPL GA trainer.
/// </summary>
/// <param name="method">The method to use.</param>
/// <param name="training">The training data to use.</param>
/// <param name="argsStr">The arguments to use.</param>
/// <returns>The newly created trainer.</returns>
public IMLTrain Create(IMLMethod method,
IMLDataSet training, String argsStr)
{
PrgPopulation pop = (PrgPopulation)method;
ICalculateScore score = new TrainingSetScore(training);
TrainEA train = new TrainEA(pop, score);
train.Rules.AddRewriteRule(new RewriteConstants());
train.Rules.AddRewriteRule(new RewriteAlgebraic());
train.CODEC = new PrgCODEC();
train.AddOperation(0.8, new SubtreeCrossover());
train.AddOperation(0.1, new SubtreeMutation(pop.Context, 4));
train.AddOperation(0.1, new ConstMutation(pop.Context, 0.5, 1.0));
train.AddScoreAdjuster(new ComplexityAdjustedScore());
train.Speciation = new PrgSpeciation();
return(train);
}
/// <summary>
/// Program entry point.
/// </summary>
/// <param name="app">Holds arguments and other info.</param>
public void Execute(IExampleInterface app)
{
IMLDataSet trainingSet = new BasicMLDataSet(XORInput, XORIdeal);
NEATPopulation pop = new NEATPopulation(2, 1, 1000);
pop.Reset();
pop.InitialConnectionDensity = 1.0; // not required, but speeds processing.
ICalculateScore score = new TrainingSetScore(trainingSet);
// train the neural network
TrainEA train = NEATUtil.ConstructNEATTrainer(pop, score);
EncogUtility.TrainToError(train, 0.01);
NEATNetwork network = (NEATNetwork)train.CODEC.Decode(train.BestGenome);
// test the neural network
Console.WriteLine(@"Neural Network Results:");
EncogUtility.Evaluate(network, trainingSet);
}
static void Main(string[] args)
{
var powerUnitRepository = new PowerUnitRepository();
var powerUnitGALogic = new PowerUnitGALogic();
var populationSize = 500;
var crossOverProbabality = 0.9;
var mutationProbabaity = 0.001;
const int MaxNumIterationsSameSolution = 25; // maximum number of iterations which GA could run on a same error rate
Display("------------ Load Power Unit Details");
var powerUnits = powerUnitRepository.PowerUnits;
DisplayPowerUnitData(powerUnits);
// 1 create initial population
Display("------------ Create Initial Population");
var population = powerUnitGALogic.CreateInitialPopulation(populationSize, powerUnits.Count);
// 2 create fitness function
Display("\n------------ Create Fitness function - sum of total diatance of cities within the chromosome - distance low --> better chromosome");
double maxPossiblePower = powerUnits.Sum(x => x.UnitCapacity);
var numberOfIntervals = new IntervalFitnessDataRepository(maxPossiblePower).GetNumberOfIntervals();
var powerUnitMaintainanceFitness = new PowerUnitMaintainanceFitnessFunction(powerUnitRepository.GetAllPowerUnits(), numberOfIntervals, maxPossiblePower);
// 3 create GA trainer
Display("\n------------ Create GA Trainer for iterations");
TrainEA geneticAlgorithm = powerUnitGALogic.CreateGA(powerUnits.Count, population, powerUnitMaintainanceFitness, crossOverProbabality, mutationProbabaity);
// 4 iterate and create off spring of new solutions
Display("\n------------ Run GA for iterations until good solutions found");
geneticAlgorithm = powerUnitGALogic.RunGA(geneticAlgorithm, MaxNumIterationsSameSolution);
// 5 display GA results
Display("\n------------ Display Final solution after iterations");
powerUnitGALogic.DisplaySolution(geneticAlgorithm, powerUnitRepository.GetAllPowerUnits(), maxPossiblePower);
Display("\n------------ Done");
Console.ReadKey();
}
static void Main(string[] args)
{
try
{
Display("------------ Load City Coodinates");
var travellingSalesman = new TravellingSalesMan();
var cities = travellingSalesman.LoadCities(Config.CityCount, Config.MaxXCordinate, Config.MaxYCordinate);
// 1 create initial population
Display("\n------------ Create Initial Solution Population");
BasicPopulation population = travellingSalesman.CreateInitialPopulation(Config.PopulationSize, Config.CityCount);
// 2 create fitness function
Display("\n------------ Create Fitness function - sum of total diatance of cities within the chromosome - distance low --> better chormosome");
var tspFitness = new TravellingSalesManFitnessScore(cities);
// 3 create GA trainer
Display("\n------------ Create GA Trainer for iterations");
TrainEA geneticAlgorithm = travellingSalesman.CreateGA(Config.CityCount, population, tspFitness, Config.CrossOverProbabality, Config.MutationProbabality);
// 4 iterate and create off spring of new solutions
Display("\n------------ Run GA for iterations until good solutions found");
geneticAlgorithm = travellingSalesman.RunGA(geneticAlgorithm, Config.MaxNumIterationsSameSolution);
// 5 display GA results
Display("\n------------ Display Final solution after iterations");
travellingSalesman.DisplaySolution(geneticAlgorithm, cities);
Display("\n------------ Fin");
Console.ReadKey();
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
/// <summary>
/// Construct a NEAT (or HyperNEAT trainer.
/// </summary>
/// <param name="population">The population.</param>
/// <param name="calculateScore">The score function.</param>
/// <returns>The NEAT EA trainer.</returns>
public static TrainEA ConstructNEATTrainer(NEATPopulation population,
ICalculateScore calculateScore)
{
var result = new TrainEA(population, calculateScore)
{
Speciation = new OriginalNEATSpeciation()
};
result.Selection = new TruncationSelection(result, 0.3);
var weightMutation = new CompoundOperator();
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(1),
new MutatePerturbLinkWeight(0.02)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(2),
new MutatePerturbLinkWeight(0.02)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(3),
new MutatePerturbLinkWeight(0.02)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectProportion(0.02),
new MutatePerturbLinkWeight(0.02)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(1),
new MutatePerturbLinkWeight(1)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(2),
new MutatePerturbLinkWeight(1)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectFixed(3),
new MutatePerturbLinkWeight(1)));
weightMutation.Components.Add(
0.1125,
new NEATMutateWeights(new SelectProportion(0.02),
new MutatePerturbLinkWeight(1)));
weightMutation.Components.Add(
0.03,
new NEATMutateWeights(new SelectFixed(1),
new MutateResetLinkWeight()));
weightMutation.Components.Add(
0.03,
new NEATMutateWeights(new SelectFixed(2),
new MutateResetLinkWeight()));
weightMutation.Components.Add(
0.03,
new NEATMutateWeights(new SelectFixed(3),
new MutateResetLinkWeight()));
weightMutation.Components.Add(
0.01,
new NEATMutateWeights(new SelectProportion(0.02),
new MutateResetLinkWeight()));
weightMutation.Components.FinalizeStructure();
result.ChampMutation = weightMutation;
result.AddOperation(0.5, new NEATCrossover());
result.AddOperation(0.494, weightMutation);
result.AddOperation(0.0005, new NEATMutateAddNode());
result.AddOperation(0.005, new NEATMutateAddLink());
result.AddOperation(0.0005, new NEATMutateRemoveLink());
result.Operators.FinalizeStructure();
if (population.IsHyperNEAT)
{
result.CODEC = new HyperNEATCODEC();
}
else
{
result.CODEC = new NEATCODEC();
}
return(result);
}
/// <summary>
/// Program entry point.
/// </summary>
/// <param name="app">Holds arguments and other info.</param>
public void Execute(IExampleInterface app)
{
IMLDataSet trainingData = GenerationUtil.GenerateSingleDataRange(
(x) => (3 * Math.Pow(x, 2) + (12 * x) + 4)
, 0, 100, 1);
EncogProgramContext context = new EncogProgramContext();
context.DefineVariable("x");
StandardExtensions.CreateNumericOperators(context);
PrgPopulation pop = new PrgPopulation(context, 1000);
MultiObjectiveFitness score = new MultiObjectiveFitness();
score.AddObjective(1.0, new TrainingSetScore(trainingData));
TrainEA genetic = new TrainEA(pop, score);
genetic.ValidationMode = true;
genetic.CODEC = new PrgCODEC();
genetic.AddOperation(0.5, new SubtreeCrossover());
genetic.AddOperation(0.25, new ConstMutation(context, 0.5, 1.0));
genetic.AddOperation(0.25, new SubtreeMutation(context, 4));
genetic.AddScoreAdjuster(new ComplexityAdjustedScore(10, 20, 10, 20.0));
genetic.Rules.AddRewriteRule(new RewriteConstants());
genetic.Rules.AddRewriteRule(new RewriteAlgebraic());
genetic.Speciation = new PrgSpeciation();
(new RampedHalfAndHalf(context, 1, 6)).Generate(new EncogRandom(), pop);
genetic.ShouldIgnoreExceptions = false;
EncogProgram best = null;
genetic.ThreadCount = 1;
try
{
for (int i = 0; i < 1000; i++)
{
genetic.Iteration();
best = (EncogProgram)genetic.BestGenome;
Console.Out.WriteLine(genetic.IterationNumber + ", Error: "
+ best.Score + ",Best Genome Size:" + best.Size
+ ",Species Count:" + pop.Species.Count + ",best: " + best.DumpAsCommonExpression());
}
//EncogUtility.evaluate(best, trainingData);
Console.Out.WriteLine("Final score:" + best.Score
+ ", effective score:" + best.AdjustedScore);
Console.Out.WriteLine(best.DumpAsCommonExpression());
//pop.dumpMembers(Integer.MAX_VALUE);
//pop.dumpMembers(10);
}
catch (Exception t)
{
Console.Out.WriteLine(t.ToString());
}
finally
{
genetic.FinishTraining();
EncogFramework.Instance.Shutdown();
}
}
public void DisplaySolution(TrainEA geneticAlgorithm, PowerUnit[] powerUnits, double maxPossiblePower)
{
try
{
var bestChromosome = (FourBitCustomGenome)geneticAlgorithm.Population.BestGenome;
FourBitGene [] bestChomosomeGenes = bestChromosome.Data;
// display best chromosome genes
for (var i = 0; i < bestChomosomeGenes.Length; i++)
{
// display chromosome to user
var geneBitsString = GetGeneBitString(bestChomosomeGenes[i]);
if (i == 0)
{
Console.Write("Best Chromosome Genes = [ {0}", geneBitsString);
}
else if (i < bestChomosomeGenes.Length - 1)
{
Console.Write(", {0}", geneBitsString);
}
else
{
Console.Write(", {0}]\n\n", geneBitsString);
}
}
var intervalFitnessDataRepository = new IntervalFitnessDataRepository(maxPossiblePower);
var intervalRawData = intervalFitnessDataRepository.IntervalRawData;
for (int i = 0; i < intervalRawData.Count; i++)
{
IntervalsFitnessData interval = intervalRawData[i];
for (int j = 0; j < bestChomosomeGenes.Length; j++)
{
PowerUnit powerUnit = powerUnits[j];
FourBitGene fourBitGene = bestChomosomeGenes[j];
int geneBitIndex = i;
var isPowerUnitMaintained = fourBitGene.Gene[geneBitIndex] == 1;
if (isPowerUnitMaintained)
{
interval.ReducedAmountOnMaintainance = interval.ReducedAmountOnMaintainance + (1 * powerUnit.UnitCapacity);
}
else
{
interval.ReducedAmountOnMaintainance = interval.ReducedAmountOnMaintainance + (0 * powerUnit.UnitCapacity);
}
}
var totalPowerReductionOnMaintanceAndUsage =
interval.PowerRequirement + interval.ReducedAmountOnMaintainance;
interval.ReserveAfterMaintainance = interval.MaxReserve - totalPowerReductionOnMaintanceAndUsage;
if (interval.ReserveAfterMaintainance < 0.0)
{
// the chromosome is not suitable for out requirement
Console.WriteLine("Error - On Interval {0} has net reserve of {1} ", interval.IntervalId, interval.ReserveAfterMaintainance);
}
}
foreach (var interval in intervalRawData)
{
Console.WriteLine(
"Interval Id = {0} , Max Reserve = {1}, Power Requirement = {2} , Reduced on maintainance = {3} , Reserve after Maintainance = {4}",
interval.IntervalId, interval.MaxReserve, interval.PowerRequirement,
interval.ReducedAmountOnMaintainance, interval.ReserveAfterMaintainance);
}
}
catch (Exception e)
{
Console.WriteLine(e);
throw;
}
}
