public static CommonOutputs.TransformOutput ConvertPredictedLabel(IHostEnvironment env, PredictedLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("ConvertPredictedLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
var predictedLabelCol = input.Data.Schema.GetColumnOrNull(input.PredictedLabelColumn);
if (!predictedLabelCol.HasValue)
{
throw host.ExceptSchemaMismatch(nameof(input), "label", input.PredictedLabelColumn);
}
var predictedLabelType = predictedLabelCol.Value.Type;
if (predictedLabelType is NumberDataViewType || predictedLabelType is BooleanDataViewType)
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, nop, input.Data), OutputData = nop
});
}
var xf = new KeyToValueMappingTransformer(host, input.PredictedLabelColumn).Transform(input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, xf, input.Data), OutputData = xf
});
}
public static CommonOutputs.TransformOutput ConvertPredictedLabel(IHostEnvironment env, PredictedLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("ConvertPredictedLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
int predictedLabelCol;
if (!input.Data.Schema.TryGetColumnIndex(input.PredictedLabelColumn, out predictedLabelCol))
{
throw host.Except($"Column '{input.PredictedLabelColumn}' not found.");
}
var predictedLabelType = input.Data.Schema.GetColumnType(predictedLabelCol);
if (predictedLabelType.IsNumber || predictedLabelType.IsBool)
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModel(env, nop, input.Data), OutputData = nop
});
}
var xf = new KeyToValueTransform(host, input.PredictedLabelColumn).Transform(input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModel(env, xf, input.Data), OutputData = xf
});
}
public static CommonOutputs.TransformOutput PrepareRegressionLabel(IHostEnvironment env, RegressionLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("PrepareRegressionLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
var labelCol = input.Data.Schema.GetColumnOrNull(input.LabelColumn);
if (!labelCol.HasValue)
{
throw host.Except($"Column '{input.LabelColumn}' not found.");
}
var labelType = labelCol.Value.Type;
if (labelType == NumberDataViewType.Single || !(labelType is NumberDataViewType))
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, nop, input.Data), OutputData = nop
});
}
var xf = new TypeConvertingTransformer(host, new TypeConvertingEstimator.ColumnOptions(input.LabelColumn, DataKind.Single, input.LabelColumn)).Transform(input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, xf, input.Data), OutputData = xf
});
}
// Factory method for SignatureDataTransform.
private static IDataTransform Create(IHostEnvironment env, Options options, IDataView input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register(RegistrationName);
host.CheckValue(options, nameof(options));
host.CheckValue(input, nameof(input));
options.Check(host);
var scores = default(VBuffer <Single>);
TrainCore(host, input, options, ref scores);
using (var ch = host.Start("Dropping Slots"))
{
int selectedCount;
var column = CreateDropSlotsColumn(options, in scores, out selectedCount);
if (column == null)
{
ch.Info("No features are being dropped.");
return(NopTransform.CreateIfNeeded(host, input));
}
ch.Info(MessageSensitivity.Schema, "Selected {0} slots out of {1} in column '{2}'", selectedCount, scores.Length, options.FeatureColumn);
return(new SlotsDroppingTransformer(host, column).Transform(input) as IDataTransform);
}
}
public static CommonOutputs.TransformOutput RenameBinaryPredictionScoreColumns(IHostEnvironment env,
RenameBinaryPredictionScoreColumnsInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("ScoreModel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
if (input.PredictorModel.Predictor.PredictionKind == PredictionKind.BinaryClassification)
{
DataViewType labelType;
var labelNames = input.PredictorModel.GetLabelInfo(host, out labelType);
if (labelNames != null && labelNames.Length == 2)
{
var positiveClass = labelNames[1];
// Rename all the score columns.
int colMax;
var maxScoreId = input.Data.Schema.GetMaxAnnotationKind(out colMax, AnnotationUtils.Kinds.ScoreColumnSetId);
var copyCols = new List <(string name, string source)>();
for (int i = 0; i < input.Data.Schema.Count; i++)
{
if (input.Data.Schema[i].IsHidden)
{
continue;
}
if (!ShouldAddColumn(input.Data.Schema, i, null, maxScoreId))
{
continue;
}
// Do not rename the PredictedLabel column.
ReadOnlyMemory <char> tmp = default;
if (input.Data.Schema.TryGetAnnotation(TextDataViewType.Instance, AnnotationUtils.Kinds.ScoreValueKind, i,
ref tmp) &&
ReadOnlyMemoryUtils.EqualsStr(AnnotationUtils.Const.ScoreValueKind.PredictedLabel, tmp))
{
continue;
}
var source = input.Data.Schema[i].Name;
var name = source + "." + positiveClass;
copyCols.Add((name, source));
}
var copyColumn = new ColumnCopyingTransformer(env, copyCols.ToArray()).Transform(input.Data);
var dropColumn = ColumnSelectingTransformer.CreateDrop(env, copyColumn, copyCols.Select(c => c.source).ToArray());
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, dropColumn, input.Data), OutputData = dropColumn
});
}
}
var newView = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, newView, input.Data), OutputData = newView
});
}
/// <summary>
/// Create method corresponding to SignatureDataTransform.
/// </summary>
public static IDataTransform Create(IHostEnvironment env, Arguments args, IDataView input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register(RegistrationName);
host.CheckValue(args, nameof(args));
host.CheckValue(input, nameof(input));
host.CheckUserArg(Utils.Size(args.Column) > 0, nameof(args.Column));
host.CheckUserArg(args.SlotsInOutput > 0, nameof(args.SlotsInOutput));
host.CheckNonWhiteSpace(args.LabelColumn, nameof(args.LabelColumn));
host.Check(args.NumBins > 1, "numBins must be greater than 1.");
using (var ch = host.Start("Selecting Slots"))
{
ch.Info("Computing mutual information");
var sw = new Stopwatch();
sw.Start();
var colSet = new HashSet <string>();
foreach (var col in args.Column)
{
if (!colSet.Add(col))
{
ch.Warning("Column '{0}' specified multiple time.", col);
}
}
var colArr = colSet.ToArray();
var colSizes = new int[colArr.Length];
var scores = MutualInformationFeatureSelectionUtils.TrainCore(host, input, args.LabelColumn, colArr,
args.NumBins, colSizes);
sw.Stop();
ch.Info("Finished mutual information computation in {0}", sw.Elapsed);
ch.Info("Selecting features to drop");
var threshold = ComputeThreshold(scores, args.SlotsInOutput, out int tiedScoresToKeep);
var columns = CreateDropSlotsColumns(colArr, colArr.Length, scores, threshold, tiedScoresToKeep, out int[] selectedCount);
if (columns.Count <= 0)
{
ch.Info("No features are being dropped.");
return(NopTransform.CreateIfNeeded(host, input));
}
for (int i = 0; i < selectedCount.Length; i++)
{
ch.Info("Selected {0} slots out of {1} in column '{2}'", selectedCount[i], colSizes[i], colArr[i]);
}
ch.Info("Total number of slots selected: {0}", selectedCount.Sum());
var dsArgs = new DropSlotsTransform.Arguments();
dsArgs.Column = columns.ToArray();
return(new DropSlotsTransform(host, dsArgs, input));
}
}
public static CommonOutputs.TransformOutput PrepareClassificationLabel(IHostEnvironment env, ClassificationLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("PrepareClassificationLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
var labelCol = input.Data.Schema.GetColumnOrNull(input.LabelColumn);
if (!labelCol.HasValue)
{
throw host.ExceptSchemaMismatch(nameof(input), "predicted label", input.LabelColumn);
}
var labelType = labelCol.Value.Type;
if (labelType is KeyType || labelType is BooleanDataViewType)
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, nop, input.Data), OutputData = nop
});
}
var args = new ValueToKeyMappingTransformer.Options()
{
Columns = new[]
{
new ValueToKeyMappingTransformer.Column()
{
Name = input.LabelColumn,
Source = input.LabelColumn,
TextKeyValues = input.TextKeyValues,
Sort = ValueToKeyMappingEstimator.KeyOrdinality.ByValue
}
}
};
var xf = ValueToKeyMappingTransformer.Create(host, args, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, xf, input.Data), OutputData = xf
});
}
public static CommonOutputs.TransformOutput PrepareClassificationLabel(IHostEnvironment env, ClassificationLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("PrepareClassificationLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
int labelCol;
if (!input.Data.Schema.TryGetColumnIndex(input.LabelColumn, out labelCol))
{
throw host.Except($"Column '{input.LabelColumn}' not found.");
}
var labelType = input.Data.Schema[labelCol].Type;
if (labelType.IsKey || labelType.IsBool)
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, nop, input.Data), OutputData = nop
});
}
var args = new ValueToKeyMappingTransformer.Arguments()
{
Column = new[]
{
new ValueToKeyMappingTransformer.Column()
{
Name = input.LabelColumn,
Source = input.LabelColumn,
TextKeyValues = input.TextKeyValues,
Sort = ValueToKeyMappingTransformer.SortOrder.Value
}
}
};
var xf = ValueToKeyMappingTransformer.Create(host, args, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, xf, input.Data), OutputData = xf
});
}
public static CommonOutputs.TransformOutput PrepareRegressionLabel(IHostEnvironment env, RegressionLabelInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("PrepareRegressionLabel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
int labelCol;
if (!input.Data.Schema.TryGetColumnIndex(input.LabelColumn, out labelCol))
{
throw host.Except($"Column '{input.LabelColumn}' not found.");
}
var labelType = input.Data.Schema[labelCol].Type;
if (labelType == NumberType.R4 || !labelType.IsNumber)
{
var nop = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, nop, input.Data), OutputData = nop
});
}
var args = new TypeConvertingTransformer.Arguments()
{
Column = new[]
{
new TypeConvertingTransformer.Column()
{
Name = input.LabelColumn,
Source = input.LabelColumn,
ResultType = DataKind.R4
}
}
};
var xf = new TypeConvertingTransformer(host, new TypeConvertingTransformer.ColumnInfo(input.LabelColumn, input.LabelColumn, DataKind.R4)).Transform(input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModelImpl(env, xf, input.Data), OutputData = xf
});
}
/// <summary>
/// Create method corresponding to SignatureDataTransform.
/// </summary>
public static IDataTransform Create(IHostEnvironment env, Arguments args, IDataView input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register(RegistrationName);
host.CheckValue(args, nameof(args));
host.CheckValue(input, nameof(input));
host.CheckUserArg(Utils.Size(args.Column) > 0, nameof(args.Column));
host.CheckUserArg(args.Count > 0, nameof(args.Count));
int[] colSizes;
var scores = CountFeatureSelectionUtils.Train(host, input, args.Column, out colSizes);
var size = args.Column.Length;
using (var ch = host.Start("Dropping Slots"))
{
int[] selectedCount;
var columns = CreateDropSlotsColumns(args, size, scores, out selectedCount);
if (columns.Count <= 0)
{
ch.Info("No features are being dropped.");
return(NopTransform.CreateIfNeeded(host, input));
}
for (int i = 0; i < selectedCount.Length; i++)
{
ch.Info(MessageSensitivity.Schema, "Selected {0} slots out of {1} in column '{2}'", selectedCount[i], colSizes[i], args.Column[i]);
}
ch.Info("Total number of slots selected: {0}", selectedCount.Sum());
var dsArgs = new DropSlotsTransform.Arguments();
dsArgs.Column = columns.ToArray();
return(new DropSlotsTransform(host, dsArgs, input));
}
}
public static CommonOutputs.TransformOutput RenameBinaryPredictionScoreColumns(IHostEnvironment env,
RenameBinaryPredictionScoreColumnsInput input)
{
Contracts.CheckValue(env, nameof(env));
var host = env.Register("ScoreModel");
host.CheckValue(input, nameof(input));
EntryPointUtils.CheckInputArgs(host, input);
if (input.PredictorModel.Predictor.PredictionKind == PredictionKind.BinaryClassification)
{
ColumnType labelType;
var labelNames = input.PredictorModel.GetLabelInfo(host, out labelType);
if (labelNames != null && labelNames.Length == 2)
{
var positiveClass = labelNames[1];
// Rename all the score columns.
int colMax;
var maxScoreId = input.Data.Schema.GetMaxMetadataKind(out colMax, MetadataUtils.Kinds.ScoreColumnSetId);
var copyCols = new List <CopyColumnsTransform.Column>();
for (int i = 0; i < input.Data.Schema.ColumnCount; i++)
{
if (input.Data.Schema.IsHidden(i))
{
continue;
}
if (!ShouldAddColumn(input.Data.Schema, i, null, maxScoreId))
{
continue;
}
// Do not rename the PredictedLabel column.
DvText tmp = default(DvText);
if (input.Data.Schema.TryGetMetadata(TextType.Instance, MetadataUtils.Kinds.ScoreValueKind, i,
ref tmp) &&
tmp.EqualsStr(MetadataUtils.Const.ScoreValueKind.PredictedLabel))
{
continue;
}
var source = input.Data.Schema.GetColumnName(i);
var name = source + "." + positiveClass;
copyCols.Add(new CopyColumnsTransform.Column()
{
Name = name, Source = source
});
}
var copyColumn = new CopyColumnsTransform(env, new CopyColumnsTransform.Arguments()
{
Column = copyCols.ToArray()
}, input.Data);
var dropColumn = new DropColumnsTransform(env, new DropColumnsTransform.Arguments()
{
Column = copyCols.Select(c => c.Source).ToArray()
}, copyColumn);
return(new CommonOutputs.TransformOutput {
Model = new TransformModel(env, dropColumn, input.Data), OutputData = dropColumn
});
}
}
var newView = NopTransform.CreateIfNeeded(env, input.Data);
return(new CommonOutputs.TransformOutput {
Model = new TransformModel(env, newView, input.Data), OutputData = newView
});
}