public void StringMatrixExtensions_ToF64Matrix()
{
var stringMatrix = new StringMatrix(InputData, 2, 3);
var actual = stringMatrix.ToF64Matrix();
Assert.AreEqual(new F64Matrix(InputData.Select(v => FloatingPointConversion.ToF64(v)).ToArray(), 2, 3), actual);
}
public void FloatingPointConversion_ToF64()
{
var value = "1.2354236";
var actual = FloatingPointConversion.ToF64(value);
Assert.AreEqual(1.2354236, actual);
}
/// <summary>
/// Parse array of feature importance values from text dump of XGBoost trees.
/// </summary>
/// <param name="textTrees"></param>
/// <param name="numberOfFeatures"></param>
/// <returns></returns>
public static double[] ParseFromTreeDump(string[] textTrees, int numberOfFeatures)
{
var importanceType = new string[] { "gain=" };
var newLine = new string[] { "\n" };
var rawFeatureImportance = new double[numberOfFeatures];
foreach (var tree in textTrees)
{
var lines = tree.Split(newLine, StringSplitOptions.RemoveEmptyEntries);
foreach (var line in lines)
{
if (!line.Contains("[") || line.Contains("booster"))
{
// Continue if line does not contain a split.
continue;
}
// Extract feature name from string between []
var name = line.Split('[')[1].Split(']')[0].Split('<')[0];
// Extract featureIndex
var featureIndex = int.Parse(name.Split('f')[1]);
// extract gain or cover
var gain = FloatingPointConversion.ToF64(line.Split(importanceType,
StringSplitOptions.RemoveEmptyEntries)[1].Split(',')[0]);
// add to featureImportance
rawFeatureImportance[featureIndex] += gain;
}
}
return(rawFeatureImportance);
}
/// <summary>
/// Parses the leaf value from a line.
/// </summary>
/// <param name="line"></param>
/// <returns></returns>
public static double ParseLeafValue(string line)
{
var valueLine = line.Split(m_leafSplit, StringSplitOptions.RemoveEmptyEntries)[1];
var valueString = valueLine.Split(',')[0];
var value = FloatingPointConversion.ToF64(valueString);
return(value);
}
/// <summary>
/// Parses the split value from a line.
/// </summary>
/// <param name="line"></param>
/// <returns></returns>
public static double ParseSplitValue(string line)
{
// Extract feature value from string between []
var valueString = line.Split('[')[1].Split(']')[0].Split('<')[1];
// convert value.
var value = FloatingPointConversion.ToF64(valueString);
return(value);
}
public void FloatingPointConversion_ToF64_to_from_custom_infinity()
{
var text = new string[] { "Inf", "beyond", "0.0", "75357.987" };
var nfi = new NumberFormatInfo()
{
PositiveInfinitySymbol = text[0],
NegativeInfinitySymbol = text[1],
};
var actual = text.Select(x => FloatingPointConversion.ToF64(x,
converter: t => double.Parse(t, nfi))).ToArray();
var expected = new double[] { double.PositiveInfinity, double.NegativeInfinity, 0.0, 75357.987 };
CollectionAssert.AreEqual(expected, actual);
}
/// <summary>
/// Converts am array of string to an array of floats
/// </summary>
/// <param name="v"></param>
/// <returns></returns>
public static double[] AsF64(this string[] v)
{
return(v.Select(s => FloatingPointConversion.ToF64(s)).ToArray());
}
static double ToF64(string value)
{
return(FloatingPointConversion.ToF64(value));
}
public void FloatingPointConversion_ToF64_Unable_To_Parse()
{
var value = "infinity12";
FloatingPointConversion.ToF64(value);
}