public void Fit(List <List <int> > X, List <int> y)
{
int numExamples = X.Count;
int numFeatures = numExamples != 0 ? X[0].Count : 0;
_stump.Fit(X, y);
if (_maxDepth <= 1 || _stump.splitVariable == -1)
{
subModel1 = null;
subModel0 = null;
return;
}
var y1 = new List <int>();
var y0 = new List <int>();
for (int i = 0; i < numExamples; i++)
{
if (X[i][_stump.splitVariable] > _stump.splitValue)
{
y1.Add(y[i]);
}
else
{
y0.Add(y[i]);
}
}
// TODO fix this to add back in randomness.
if (y1.Count != 0)
{
subModel1 = new DecisionTree(_maxDepth - 1, new DecisionStumpInfoGain());
subModel1.Fit(X.Where(x => x[_stump.splitVariable] > _stump.splitValue).ToList(), y1);
}
else
{
subModel1 = null;
}
if (y0.Count != 0)
{
subModel0 = new DecisionTree(_maxDepth - 1, new DecisionStumpInfoGain());
subModel0.Fit(X.Where(x => x[_stump.splitVariable] <= _stump.splitValue).ToList(), y0);
}
else
{
subModel0 = null;
}
}
public void Fit(List <List <int> > X, List <int> y)
{
// TODO this is duplicated.
int numExamples = X.Count;
var bootstrapX = new List <List <int> >();
var bootstrapY = new List <int>();
// Fill up array with random examples with replacement.
var rand = new Random();
for (int i = 0; i < numExamples; i++)
{
var idx = rand.Next(0, numExamples);
bootstrapX.Add(X[idx]);
bootstrapY.Add(y[idx]);
}
decisionTreeRoot.Fit(bootstrapX, bootstrapY);
}