protected override ISymbolicExpressionTree OptimizeConstants(ISymbolicExpressionTree tree, IProgress progress)
{
const int constOptIterations = 50;
const int maxRepetitions = 1000;
var regressionProblemData = Content.ProblemData;
var model = Content.Model;
progress.CanBeStopped = true;
var prevResult = 0.0;
var result = 0.0;
int reps = 0;
do
{
prevResult = result;
result = SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(model.Interpreter, tree, regressionProblemData, regressionProblemData.TrainingIndices,
applyLinearScaling: true, maxIterations: constOptIterations, updateVariableWeights: true, lowerEstimationLimit: model.LowerEstimationLimit, upperEstimationLimit: model.UpperEstimationLimit,
iterationCallback: (args, func, obj) => {
double newProgressValue = progress.ProgressValue + 1.0 / (constOptIterations + 2); // (maxIterations + 2) iterations are reported
progress.ProgressValue = Math.Min(newProgressValue, 1.0);
});
reps++;
} while (prevResult < result && reps < maxRepetitions &&
progress.ProgressState != ProgressState.StopRequested &&
progress.ProgressState != ProgressState.CancelRequested);
return(tree);
}
public static ISymbolicRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData, string modelStructure, int maxIterations)
{
var parser = new InfixExpressionParser();
var tree = parser.Parse(modelStructure);
var simplifier = new SymbolicDataAnalysisExpressionTreeSimplifier();
if (!SymbolicRegressionConstantOptimizationEvaluator.CanOptimizeConstants(tree))
{
throw new ArgumentException("The optimizer does not support the specified model structure.");
}
var interpreter = new SymbolicDataAnalysisExpressionTreeLinearInterpreter();
SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(interpreter, tree, problemData, problemData.TrainingIndices,
applyLinearScaling: false, maxIterations: maxIterations,
updateVariableWeights: false, updateConstantsInTree: true);
var scaledModel = new SymbolicRegressionModel(problemData.TargetVariable, tree, (ISymbolicDataAnalysisExpressionTreeInterpreter)interpreter.Clone());
scaledModel.Scale(problemData);
SymbolicRegressionSolution solution = new SymbolicRegressionSolution(scaledModel, (IRegressionProblemData)problemData.Clone());
solution.Model.Name = "Regression Model";
solution.Name = "Regression Solution";
return(solution);
}
protected override void btnOptimizeConstants_Click(object sender, EventArgs e)
{
var model = Content.Model;
SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(model.Interpreter, model.SymbolicExpressionTree, Content.ProblemData, Content.ProblemData.TrainingIndices,
applyLinearScaling: true, maxIterations: 50, updateVariableWeights: true, lowerEstimationLimit: model.LowerEstimationLimit, upperEstimationLimit: model.UpperEstimationLimit);
UpdateModel(Content.Model.SymbolicExpressionTree);
}
protected override ISymbolicExpressionTree OptimizeConstants(ISymbolicExpressionTree tree, IProgress progress)
{
const int constOptIterations = 50;
var regressionProblemData = Content.ProblemData;
var model = Content.Model;
SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(model.Interpreter, tree, regressionProblemData, regressionProblemData.TrainingIndices,
applyLinearScaling: true, maxIterations: constOptIterations, updateVariableWeights: true, lowerEstimationLimit: model.LowerEstimationLimit, upperEstimationLimit: model.UpperEstimationLimit,
iterationCallback: (args, func, obj) => {
double newProgressValue = progress.ProgressValue + 1.0 / (constOptIterations + 2); // (maxIterations + 2) iterations are reported
progress.ProgressValue = Math.Min(newProgressValue, 1.0);
});
return(tree);
}
private OffspringSelectionGeneticAlgorithm CreateGpSymbolicRegressionSample()
{
var osga = new OffspringSelectionGeneticAlgorithm();
#region Problem Configuration
var provider = new VariousInstanceProvider(seed);
var instance = provider.GetDataDescriptors().First(x => x.Name.StartsWith("Spatial co-evolution"));
var problemData = (RegressionProblemData)provider.LoadData(instance);
var problem = new SymbolicRegressionSingleObjectiveProblem();
problem.ProblemData = problemData;
problem.Load(problemData);
problem.BestKnownQuality.Value = 1.0;
#region configure grammar
var grammar = (TypeCoherentExpressionGrammar)problem.SymbolicExpressionTreeGrammar;
grammar.ConfigureAsDefaultRegressionGrammar();
//symbols square, power, squareroot, root, log, exp, sine, cosine, tangent, variable
var square = grammar.Symbols.OfType <Square>().Single();
var power = grammar.Symbols.OfType <Power>().Single();
var squareroot = grammar.Symbols.OfType <SquareRoot>().Single();
var root = grammar.Symbols.OfType <Root>().Single();
var log = grammar.Symbols.OfType <Logarithm>().Single();
var exp = grammar.Symbols.OfType <Exponential>().Single();
var sine = grammar.Symbols.OfType <Sine>().Single();
var cosine = grammar.Symbols.OfType <Cosine>().Single();
var tangent = grammar.Symbols.OfType <Tangent>().Single();
var variable = grammar.Symbols.OfType <Variable>().Single();
var powerSymbols = grammar.Symbols.Single(s => s.Name == "Power Functions");
powerSymbols.Enabled = true;
square.Enabled = true;
square.InitialFrequency = 1.0;
foreach (var allowed in grammar.GetAllowedChildSymbols(square))
{
grammar.RemoveAllowedChildSymbol(square, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(square, 0))
{
grammar.RemoveAllowedChildSymbol(square, allowed, 0);
}
grammar.AddAllowedChildSymbol(square, variable);
power.Enabled = false;
squareroot.Enabled = false;
foreach (var allowed in grammar.GetAllowedChildSymbols(squareroot))
{
grammar.RemoveAllowedChildSymbol(squareroot, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(squareroot, 0))
{
grammar.RemoveAllowedChildSymbol(squareroot, allowed, 0);
}
grammar.AddAllowedChildSymbol(squareroot, variable);
root.Enabled = false;
log.Enabled = true;
log.InitialFrequency = 1.0;
foreach (var allowed in grammar.GetAllowedChildSymbols(log))
{
grammar.RemoveAllowedChildSymbol(log, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(log, 0))
{
grammar.RemoveAllowedChildSymbol(log, allowed, 0);
}
grammar.AddAllowedChildSymbol(log, variable);
exp.Enabled = true;
exp.InitialFrequency = 1.0;
foreach (var allowed in grammar.GetAllowedChildSymbols(exp))
{
grammar.RemoveAllowedChildSymbol(exp, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(exp, 0))
{
grammar.RemoveAllowedChildSymbol(exp, allowed, 0);
}
grammar.AddAllowedChildSymbol(exp, variable);
sine.Enabled = false;
foreach (var allowed in grammar.GetAllowedChildSymbols(sine))
{
grammar.RemoveAllowedChildSymbol(sine, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(sine, 0))
{
grammar.RemoveAllowedChildSymbol(sine, allowed, 0);
}
grammar.AddAllowedChildSymbol(sine, variable);
cosine.Enabled = false;
foreach (var allowed in grammar.GetAllowedChildSymbols(cosine))
{
grammar.RemoveAllowedChildSymbol(cosine, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(cosine, 0))
{
grammar.RemoveAllowedChildSymbol(cosine, allowed, 0);
}
grammar.AddAllowedChildSymbol(cosine, variable);
tangent.Enabled = false;
foreach (var allowed in grammar.GetAllowedChildSymbols(tangent))
{
grammar.RemoveAllowedChildSymbol(tangent, allowed);
}
foreach (var allowed in grammar.GetAllowedChildSymbols(tangent, 0))
{
grammar.RemoveAllowedChildSymbol(tangent, allowed, 0);
}
grammar.AddAllowedChildSymbol(tangent, variable);
#endregion
problem.SymbolicExpressionTreeGrammar = grammar;
// configure remaining problem parameters
problem.MaximumSymbolicExpressionTreeLength.Value = 50;
problem.MaximumSymbolicExpressionTreeDepth.Value = 12;
problem.MaximumFunctionDefinitions.Value = 0;
problem.MaximumFunctionArguments.Value = 0;
var evaluator = new SymbolicRegressionConstantOptimizationEvaluator();
evaluator.ConstantOptimizationIterations.Value = 5;
problem.EvaluatorParameter.Value = evaluator;
problem.RelativeNumberOfEvaluatedSamplesParameter.Hidden = true;
problem.SolutionCreatorParameter.Hidden = true;
#endregion
#region Algorithm Configuration
osga.Problem = problem;
osga.Name = "Offspring Selection Genetic Programming - Symbolic Regression";
osga.Description = "Genetic programming with strict offspring selection for solving a benchmark regression problem.";
SamplesUtils.ConfigureOsGeneticAlgorithmParameters <GenderSpecificSelector, SubtreeCrossover, MultiSymbolicExpressionTreeManipulator>(osga, 100, 1, 25, 0.2, 50);
var mutator = (MultiSymbolicExpressionTreeManipulator)osga.Mutator;
mutator.Operators.OfType <FullTreeShaker>().Single().ShakingFactor = 0.1;
mutator.Operators.OfType <OnePointShaker>().Single().ShakingFactor = 1.0;
osga.Analyzer.Operators.SetItemCheckedState(
osga.Analyzer.Operators
.OfType <SymbolicRegressionSingleObjectiveOverfittingAnalyzer>()
.Single(), false);
osga.Analyzer.Operators.SetItemCheckedState(
osga.Analyzer.Operators
.OfType <SymbolicDataAnalysisAlleleFrequencyAnalyzer>()
.First(), false);
osga.ComparisonFactorModifierParameter.Hidden = true;
osga.ComparisonFactorLowerBoundParameter.Hidden = true;
osga.ComparisonFactorUpperBoundParameter.Hidden = true;
osga.OffspringSelectionBeforeMutationParameter.Hidden = true;
osga.SuccessRatioParameter.Hidden = true;
osga.SelectedParentsParameter.Hidden = true;
osga.ElitesParameter.Hidden = true;
#endregion
return(osga);
}
/// <summary>
/// Fits a model to the data by optimizing the numeric constants.
/// Model is specified as infix expression containing variable names and numbers.
/// The starting point for the numeric constants is initialized randomly if a random number generator is specified (~N(0,1)). Otherwise the user specified constants are
/// used as a starting point.
/// </summary>-
/// <param name="problemData">Training and test data</param>
/// <param name="modelStructure">The function as infix expression</param>
/// <param name="maxIterations">Number of constant optimization iterations (using Levenberg-Marquardt algorithm)</param>
/// <param name="random">Optional random number generator for random initialization of numeric constants.</param>
/// <returns></returns>
public static ISymbolicRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData, string modelStructure, int maxIterations, bool applyLinearScaling, IRandom rand = null)
{
var parser = new InfixExpressionParser();
var tree = parser.Parse(modelStructure);
// parser handles double and string variables equally by creating a VariableTreeNode
// post-process to replace VariableTreeNodes by FactorVariableTreeNodes for all string variables
var factorSymbol = new FactorVariable();
factorSymbol.VariableNames =
problemData.AllowedInputVariables.Where(name => problemData.Dataset.VariableHasType <string>(name));
factorSymbol.AllVariableNames = factorSymbol.VariableNames;
factorSymbol.VariableValues =
factorSymbol.VariableNames.Select(name =>
new KeyValuePair <string, Dictionary <string, int> >(name,
problemData.Dataset.GetReadOnlyStringValues(name).Distinct()
.Select((n, i) => Tuple.Create(n, i))
.ToDictionary(tup => tup.Item1, tup => tup.Item2)));
foreach (var parent in tree.IterateNodesPrefix().ToArray())
{
for (int i = 0; i < parent.SubtreeCount; i++)
{
var varChild = parent.GetSubtree(i) as VariableTreeNode;
var factorVarChild = parent.GetSubtree(i) as FactorVariableTreeNode;
if (varChild != null && factorSymbol.VariableNames.Contains(varChild.VariableName))
{
parent.RemoveSubtree(i);
var factorTreeNode = (FactorVariableTreeNode)factorSymbol.CreateTreeNode();
factorTreeNode.VariableName = varChild.VariableName;
factorTreeNode.Weights =
factorTreeNode.Symbol.GetVariableValues(factorTreeNode.VariableName).Select(_ => 1.0).ToArray();
// weight = 1.0 for each value
parent.InsertSubtree(i, factorTreeNode);
}
else if (factorVarChild != null && factorSymbol.VariableNames.Contains(factorVarChild.VariableName))
{
if (factorSymbol.GetVariableValues(factorVarChild.VariableName).Count() != factorVarChild.Weights.Length)
{
throw new ArgumentException(
string.Format("Factor variable {0} needs exactly {1} weights",
factorVarChild.VariableName,
factorSymbol.GetVariableValues(factorVarChild.VariableName).Count()));
}
parent.RemoveSubtree(i);
var factorTreeNode = (FactorVariableTreeNode)factorSymbol.CreateTreeNode();
factorTreeNode.VariableName = factorVarChild.VariableName;
factorTreeNode.Weights = factorVarChild.Weights;
parent.InsertSubtree(i, factorTreeNode);
}
}
}
if (!SymbolicRegressionConstantOptimizationEvaluator.CanOptimizeConstants(tree))
{
throw new ArgumentException("The optimizer does not support the specified model structure.");
}
// initialize constants randomly
if (rand != null)
{
foreach (var node in tree.IterateNodesPrefix().OfType <ConstantTreeNode>())
{
double f = Math.Exp(NormalDistributedRandom.NextDouble(rand, 0, 1));
double s = rand.NextDouble() < 0.5 ? -1 : 1;
node.Value = s * node.Value * f;
}
}
var interpreter = new SymbolicDataAnalysisExpressionTreeLinearInterpreter();
SymbolicRegressionConstantOptimizationEvaluator.OptimizeConstants(interpreter, tree, problemData, problemData.TrainingIndices,
applyLinearScaling: applyLinearScaling, maxIterations: maxIterations,
updateVariableWeights: false, updateConstantsInTree: true);
var model = new SymbolicRegressionModel(problemData.TargetVariable, tree, (ISymbolicDataAnalysisExpressionTreeInterpreter)interpreter.Clone());
if (applyLinearScaling)
{
model.Scale(problemData);
}
SymbolicRegressionSolution solution = new SymbolicRegressionSolution(model, (IRegressionProblemData)problemData.Clone());
solution.Model.Name = "Regression Model";
solution.Name = "Regression Solution";
return(solution);
}
