public override void SaveSummary(TextWriter writer, RoleMappedSchema schema)
{
var names = default(VBuffer <DvText>);
MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, Weight.Length, ref names);
writer.WriteLine("Ordinary Least Squares Model Summary");
writer.WriteLine("R-squared: {0:g4}", RSquared);
if (HasStatistics)
{
writer.WriteLine("Adjusted R-squared: {0:g4}", RSquaredAdjusted);
writer.WriteLine();
writer.WriteLine("Index\tName\tWeight\tStdErr\tt-Value\tp-Value");
const string format = "{0}\t{1}\t{2}\t{3:g4}\t{4:g4}\t{5:e4}";
writer.WriteLine(format, "", "Bias", Bias, _standardErrors[0], _tValues[0], _pValues[0]);
Contracts.Assert(Weight.IsDense);
var coeffs = Weight.Values;
for (int i = 0; i < coeffs.Length; i++)
{
var name = names.GetItemOrDefault(i);
writer.WriteLine(format, i, DvText.Identical(name, DvText.Empty) ? $"f{i}" : name.ToString(),
coeffs[i], _standardErrors[i + 1], _tValues[i + 1], _pValues[i + 1]);
}
}
else
{
writer.WriteLine();
writer.WriteLine("Index\tName\tWeight");
const string format = "{0}\t{1}\t{2}";
writer.WriteLine(format, "", "Bias", Bias);
Contracts.Assert(Weight.IsDense);
var coeffs = Weight.Values;
for (int i = 0; i < coeffs.Length; i++)
{
var name = names.GetItemOrDefault(i);
writer.WriteLine(format, i, DvText.Identical(name, DvText.Empty) ? $"f{i}" : name.ToString(), coeffs[i]);
}
}
}
public static IEnumerable <KeyValuePair <string, Single> > GetSortedLinearModelFeatureNamesAndWeights(Single bias,
ref VBuffer <Single> weights, ref VBuffer <DvText> names)
{
var orderedWeights = weights.Items()
.Where(weight => Math.Abs(weight.Value) >= Epsilon)
.OrderByDescending(kv => Math.Abs(kv.Value));
var list = new List <KeyValuePair <string, Single> >()
{
new KeyValuePair <string, Single>("(Bias)", bias)
};
foreach (var weight in orderedWeights)
{
int index = weight.Key;
var name = names.GetItemOrDefault(index);
list.Add(new KeyValuePair <string, Single>(
DvText.Identical(name, DvText.Empty) ? $"f{index}" : name.ToString(), weight.Value));
}
return(list);
}
///<inheritdoc/>
public IList <KeyValuePair <string, object> > GetSummaryInKeyValuePairs(RoleMappedSchema schema)
{
Host.CheckValueOrNull(schema);
List <KeyValuePair <string, object> > results = new List <KeyValuePair <string, object> >();
var names = default(VBuffer <DvText>);
MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, _numFeatures, ref names);
for (int classNumber = 0; classNumber < _biases.Length; classNumber++)
{
results.Add(new KeyValuePair <string, object>(
string.Format("{0}+(Bias)", GetLabelName(classNumber)),
_biases[classNumber]
));
}
for (int classNumber = 0; classNumber < _weights.Length; classNumber++)
{
var orderedWeights = _weights[classNumber].Items().OrderByDescending(kv => Math.Abs(kv.Value));
foreach (var weight in orderedWeights)
{
var value = weight.Value;
if (value == 0)
{
break;
}
int index = weight.Key;
var name = names.GetItemOrDefault(index);
results.Add(new KeyValuePair <string, object>(
string.Format("{0}+{1}", GetLabelName(classNumber), DvText.Identical(name, DvText.Empty) ? $"f{index}" : name.ToString()),
value
));
}
}
return(results);
}
/// <summary>
/// Build a Bing TreeEnsemble .ini representation of the given predictor
/// </summary>
public static string LinearModelAsIni(ref VBuffer <Float> weights, Float bias, IPredictor predictor = null,
RoleMappedSchema schema = null, PlattCalibrator calibrator = null)
{
// TODO: Might need to consider a max line length for the Weights list, requiring us to split it up into
// multiple evaluators
StringBuilder inputBuilder = new StringBuilder();
StringBuilder aggregatedNodesBuilder = new StringBuilder("Nodes=");
StringBuilder weightsBuilder = new StringBuilder("Weights=");
var featureNames = default(VBuffer <DvText>);
MetadataUtils.GetSlotNames(schema, RoleMappedSchema.ColumnRole.Feature, weights.Length, ref featureNames);
int numNonZeroWeights = 0;
const string weightsSep = "\t";
VBufferUtils.ForEachDefined(ref weights,
(idx, value) =>
{
if (Math.Abs(value - 0) >= Epsilon)
{
numNonZeroWeights++;
var name = featureNames.GetItemOrDefault(idx);
inputBuilder.AppendLine("[Input:" + numNonZeroWeights + "]");
inputBuilder.AppendLine("Name=" + (featureNames.Count == 0 ? "Feature_" + idx : DvText.Identical(name, DvText.Empty) ? $"f{idx}" : name.ToString()));
inputBuilder.AppendLine("Transform=linear");
inputBuilder.AppendLine("Slope=1");
inputBuilder.AppendLine("Intercept=0");
inputBuilder.AppendLine();
aggregatedNodesBuilder.Append("I:" + numNonZeroWeights + weightsSep);
weightsBuilder.Append(value + weightsSep);
}
});
StringBuilder builder = new StringBuilder();
builder.AppendLine("[TreeEnsemble]");
builder.AppendLine("Inputs=" + numNonZeroWeights);
builder.AppendLine("Evaluators=1");
builder.AppendLine();
builder.AppendLine(inputBuilder.ToString());
builder.AppendLine("[Evaluator:1]");
builder.AppendLine("EvaluatorType=Aggregator");
builder.AppendLine("Type=Linear");
builder.AppendLine("Bias=" + bias);
builder.AppendLine("NumNodes=" + numNonZeroWeights);
builder.AppendLine(aggregatedNodesBuilder.ToString().Trim());
builder.AppendLine(weightsBuilder.ToString().Trim());
#if false // REVIEW: This should be done by the caller using the actual training args!
builder.AppendLine();
builder.AppendLine("[Comments]");
builder.Append("Trained by TLC");
if (predictor != null)
{
builder.Append(" as /cl " + predictor.GetType().Name);
if (predictor is IInitializable)
{
string settings = string.Join(";", (predictor as IInitializable).GetSettings());
if (!string.IsNullOrEmpty(settings))
{
builder.Append(" /cls " + settings);
}
}
}
#endif
string ini = builder.ToString();
// Add the calibration if the model was trained with calibration
if (calibrator != null)
{
string calibratorEvaluatorIni = IniFileUtils.GetCalibratorEvaluatorIni(ini, calibrator);
ini = IniFileUtils.AddEvaluator(ini, calibratorEvaluatorIni);
}
return(ini);
}