public void PruningState_should_merge_phase_correctly()
{
var p1 = new PruningState(_node1, PruningInitialized.Empty);
var p2 = new PruningState(_node1, PruningPerformed.Instance);
Assert.Equal(PruningPerformed.Instance, p1.Merge(p2).Phase);
Assert.Equal(PruningPerformed.Instance, p2.Merge(p1).Phase);
}
public void PruningState_should_merge_seen_correctly()
{
var p1 = new PruningState(_node1, new PruningInitialized(_node2.Address));
var p2 = new PruningState(_node1, new PruningInitialized(_node4.Address));
var expected = new PruningState(_node1, new PruningInitialized(_node2.Address, _node4.Address));
Assert.Equal(expected, p1.Merge(p2));
Assert.Equal(expected, p2.Merge(p1));
}
public void PruningState_should_merge_owner_correctly()
{
var p1 = new PruningState(_node1, PruningInitialized.Empty);
var p2 = new PruningState(_node2, PruningInitialized.Empty);
var expected = new PruningState(_node1, PruningInitialized.Empty);
Assert.Equal(expected, p1.Merge(p2));
Assert.Equal(expected, p2.Merge(p1));
}
protected PruningState(PruningState original, Cloner cloner) : base(original, cloner)
{
modelComplexities = original.modelComplexities.ToDictionary(x => cloner.Clone(x.Key), x => x.Value);
nodeComplexities = original.nodeComplexities.ToDictionary(x => cloner.Clone(x.Key), x => x.Value);
pruningSizes = original.pruningSizes.ToDictionary(x => cloner.Clone(x.Key), x => x.Value);
modelErrors = original.modelErrors.ToDictionary(x => cloner.Clone(x.Key), x => x.Value);
nodeQueue = new Queue <RegressionNodeModel>(original.nodeQueue.Select(cloner.Clone));
trainingRowsQueue = new Queue <IReadOnlyList <int> >(original.trainingRowsQueue.Select(x => (IReadOnlyList <int>)x.ToArray()));
pruningRowsQueue = new Queue <IReadOnlyList <int> >(original.pruningRowsQueue.Select(x => (IReadOnlyList <int>)x.ToArray()));
Code = original.Code;
}
private static void Prune(RegressionNodeTreeModel tree, PruningState state, double pruningStrength)
{
if (state.Code == 3)
{
state.FillTopDown(tree);
state.Code = 4;
}
while (state.nodeQueue.Count != 0)
{
var n = state.nodeQueue.Dequeue();
if (n.IsLeaf || pruningStrength <= n.PruningStrength)
{
continue;
}
n.ToLeaf();
}
}
private static void UpdateThreshold(RegressionNodeTreeModel tree, PruningState state)
{
if (state.Code == 2)
{
state.FillTopDown(tree);
state.Code = 3;
}
while (state.nodeQueue.Count != 0)
{
var n = state.nodeQueue.Dequeue();
if (n.IsLeaf || n.Parent == null || double.IsNaN(n.Parent.PruningStrength))
{
continue;
}
n.PruningStrength = Math.Min(n.PruningStrength, n.Parent.PruningStrength);
}
}
private static void AssignPruningThresholds(RegressionNodeTreeModel tree, PruningState state, double pruningDecay)
{
if (state.Code == 1)
{
state.FillBottomUp(tree);
state.Code = 2;
}
while (state.nodeQueue.Count != 0)
{
var n = state.nodeQueue.Dequeue();
if (n.IsLeaf)
{
continue;
}
n.PruningStrength = PruningThreshold(state.pruningSizes[n], state.modelComplexities[n], state.nodeComplexities[n], state.modelErrors[n], SubtreeError(n, state.pruningSizes, state.modelErrors), pruningDecay);
}
}
private static void InstallModels(RegressionNodeTreeModel tree, PruningState state, IReadOnlyList <int> trainingRows, IReadOnlyList <int> pruningRows, ILeafModel leaf, RegressionTreeParameters regressionTreeParams, CancellationToken cancellationToken)
{
if (state.Code == 0)
{
state.FillBottomUp(tree, trainingRows, pruningRows, regressionTreeParams.Data);
state.Code = 1;
}
while (state.nodeQueue.Count != 0)
{
cancellationToken.ThrowIfCancellationRequested();
var n = state.nodeQueue.Peek();
var training = state.trainingRowsQueue.Peek();
var pruning = state.pruningRowsQueue.Peek();
BuildPruningModel(n, leaf, training, pruning, state, regressionTreeParams, cancellationToken);
state.nodeQueue.Dequeue();
state.trainingRowsQueue.Dequeue();
state.pruningRowsQueue.Dequeue();
}
}
private static void BuildPruningModel(RegressionNodeModel regressionNode, ILeafModel leaf, IReadOnlyList <int> trainingRows, IReadOnlyList <int> pruningRows, PruningState state, RegressionTreeParameters regressionTreeParams, CancellationToken cancellationToken)
{
//create regressionProblemdata from pruning data
var vars = regressionTreeParams.AllowedInputVariables.Concat(new[] { regressionTreeParams.TargetVariable }).ToArray();
var reducedData = new Dataset(vars, vars.Select(x => regressionTreeParams.Data.GetDoubleValues(x, pruningRows).ToList()));
var pd = new RegressionProblemData(reducedData, regressionTreeParams.AllowedInputVariables, regressionTreeParams.TargetVariable);
pd.TrainingPartition.Start = pd.TrainingPartition.End = pd.TestPartition.Start = 0;
pd.TestPartition.End = reducedData.Rows;
//build pruning model
int numModelParams;
var model = leaf.BuildModel(trainingRows, regressionTreeParams, cancellationToken, out numModelParams);
//record error and complexities
var rmsModel = model.CreateRegressionSolution(pd).TestRootMeanSquaredError;
state.pruningSizes.Add(regressionNode, pruningRows.Count);
state.modelErrors.Add(regressionNode, rmsModel);
state.modelComplexities.Add(regressionNode, numModelParams);
if (regressionNode.IsLeaf)
{
state.nodeComplexities[regressionNode] = state.modelComplexities[regressionNode];
}
else
{
state.nodeComplexities.Add(regressionNode, state.nodeComplexities[regressionNode.Left] + state.nodeComplexities[regressionNode.Right] + 1);
}
}
