private void AreEqual(Node n1, Node n2, bool ignoreParent)
{
Assert.AreEqual(n1.IsLeaf, n2.IsLeaf);
Assert.AreEqual(n1.Value, n2.Value);
Assert.AreEqual(n1.Label, n2.Label);
Assert.AreEqual(n1.Column, n2.Column);
Assert.AreEqual(n1.Name, n2.Name);
if (n1.Edges == null) Assert.IsNull(n2.Edges);
else
{
Assert.AreEqual(n1.Edges.Length, n2.Edges.Length);
// since we are not ignoring parent,
// no need to check edges since these
// are checked when ignoring parents
if (ignoreParent) return;
for (int i = 0; i < n1.Edges.Length; i++)
{
var n1e = n1.Edges[i];
var n2e = n2.Edges[i];
Assert.AreEqual(n1e.Min, n2e.Min);
Assert.AreEqual(n1e.Max, n2e.Max);
Assert.AreEqual(n1e.Discrete, n2e.Discrete);
Assert.AreEqual(n1e.Label, n2e.Label);
if (!ignoreParent)
AreEqual(n1e.Parent, n2e.Parent, true);
AreEqual(n1e.Child, n2e.Child, true);
}
}
}
private double WalkNode(Vector v, Node node)
{
if (node.IsLeaf)
return node.Value;
// Get the index of the feature for this node.
var col = node.Column;
if (col == -1)
throw new InvalidOperationException("Invalid Feature encountered during node walk!");
for (int i = 0; i < node.Edges.Length; i++)
{
Edge edge = node.Edges[i];
if (edge.Discrete && v[col] == edge.Min)
return WalkNode(v, edge.Child);
if (!edge.Discrete && v[col] >= edge.Min && v[col] < edge.Max)
return WalkNode(v, edge.Child);
}
if (Hint != double.Epsilon)
return Hint;
else
throw new InvalidOperationException(String.Format("Unable to match split value {0} for feature {1}[2]\nConsider setting a Hint in order to avoid this error.", v[col], Descriptor.At(col), col));
}
private string PrintNode(Node n, string pre)
{
if (n.IsLeaf)
return String.Format("{0} +({1}, {2:#.####})\n", pre, n.Label, n.Value);
else
{
StringBuilder sb = new StringBuilder();
sb.AppendLine(String.Format("{0}[{1}, {2:0.0000}]", pre, n.Name, n.Gain));
foreach (Edge edge in n.Edges)
{
sb.AppendLine(String.Format("{0} |- {1}", pre, edge.Label));
sb.Append(PrintNode(edge.Child, String.Format("{0} |\t", pre)));
}
return sb.ToString();
}
}
private Node BuildTree(Matrix x, Vector y, int depth, List<int> used)
{
if (depth < 0)
return BuildLeafNode(y.Mode());
var tuple = GetBestSplit(x, y, used);
var col = tuple.Item1;
var gain = tuple.Item2;
var measure = tuple.Item3;
// uh oh, need to return something?
// a weird node of some sort...
// but just in case...
if (col == -1)
return BuildLeafNode(y.Mode());
used.Add(col);
Node node = new Node
{
Column = col,
Gain = gain,
IsLeaf = false,
Name = Descriptor.ColumnAt(col)
};
// populate edges
List<Edge> edges = new List<Edge>(measure.Segments.Length);
for (int i = 0; i < measure.Segments.Length; i++)
{
// working set
var segment = measure.Segments[i];
var edge = new Edge()
{
Parent = node,
Discrete = measure.Discrete,
Min = segment.Min,
Max = segment.Max
};
IEnumerable<int> slice;
if (edge.Discrete)
{
// get discrete label
edge.Label = Descriptor.At(col).Convert(segment.Min).ToString();
// do value check for matrix slicing
slice = x.Indices(v => v[col] == segment.Min);
}
else
{
// get range label
edge.Label = string.Format("{0} ≤ x < {1}", segment.Min, segment.Max);
// do range check for matrix slicing
slice = x.Indices(v => v[col] >= segment.Min && v[col] < segment.Max);
}
// something to look at?
// if this number is 0 then this edge
// leads to a dead end - the edge will
// not be built
if (slice.Count() > 0)
{
Vector ySlice = y.Slice(slice);
// only one answer, set leaf
if (ySlice.Distinct().Count() == 1)
edge.Child = BuildLeafNode(ySlice[0]);
// otherwise continue to build tree
else
edge.Child = BuildTree(x.Slice(slice), ySlice, depth - 1, used);
edges.Add(edge);
}
}
// might check if there are no edges
// if this is the case should convert
// node to leaf and bail
var egs = edges.ToArray();
// problem, need to convert
// parent to terminal node
// with mode
if (egs.Length <= 1)
node = BuildLeafNode(y.Mode());
else
node.Edges = egs;
return node;
}
private void ReLinkNodes(Node n)
{
if (n.Edges != null)
{
foreach (Edge e in n.Edges)
{
e.Parent = n;
if (e.Child.IsLeaf)
e.Child.Label = Descriptor.Label.Convert(e.Child.Value);
else
ReLinkNodes(e.Child);
}
}
}
public void ReadXml(XmlReader reader)
{
reader.MoveToContent();
Hint = double.Parse(reader.GetAttribute("Hint"));
reader.ReadStartElement();
XmlSerializer dserializer = new XmlSerializer(typeof(Descriptor));
Descriptor = (Descriptor)dserializer.Deserialize(reader);
reader.Read();
XmlSerializer serializer = new XmlSerializer(typeof(Node));
Tree = (Node)serializer.Deserialize(reader);
// re-establish tree cycles and values
ReLinkNodes(Tree);
}