/// <summary>
/// Returns the <see cref="SchemaShape"/> of the schema which will be produced by the transformer.
/// Used for schema propagation and verification in a pipeline.
/// </summary>
public SchemaShape GetOutputSchema(SchemaShape inputSchema)
{
_host.CheckValue(inputSchema, nameof(inputSchema));
var result = inputSchema.ToDictionary(x => x.Name);
foreach (var colPair in _columns)
{
if (!inputSchema.TryFindColumn(colPair.InputColumnName, out var col))
{
throw _host.ExceptSchemaMismatch(nameof(inputSchema), "input", colPair.InputColumnName);
}
if (!CountFeatureSelectionUtils.IsValidColumnType(col.ItemType))
{
throw _host.ExceptUserArg(nameof(inputSchema), "Column '{0}' does not have compatible type. Expected types are float, double or string.", colPair.InputColumnName);
}
var metadata = new List <SchemaShape.Column>();
if (col.Annotations.TryFindColumn(AnnotationUtils.Kinds.SlotNames, out var slotMeta))
{
metadata.Add(slotMeta);
}
if (col.Annotations.TryFindColumn(AnnotationUtils.Kinds.CategoricalSlotRanges, out var categoricalSlotMeta))
{
metadata.Add(categoricalSlotMeta);
}
if (col.IsNormalized() && col.Annotations.TryFindColumn(AnnotationUtils.Kinds.IsNormalized, out var isNormalizedAnnotation))
{
metadata.Add(isNormalizedAnnotation);
}
result[colPair.Name] = new SchemaShape.Column(colPair.Name, col.Kind, col.ItemType, false, new SchemaShape(metadata.ToArray()));
}
return(new SchemaShape(result.Values));
}
/// <summary>
/// Trains and returns a <see cref="ITransformer"/>.
/// </summary>
public ITransformer Fit(IDataView input)
{
_host.CheckValue(input, nameof(input));
int[] colSizes;
var scores = CountFeatureSelectionUtils.Train(_host, input, _columns.Select(column => column.InputColumnName).ToArray(), out colSizes);
var size = _columns.Length;
using (var ch = _host.Start("Dropping Slots"))
{
// If no slots should be dropped from a column, use copy column to generate the corresponding output column.
SlotsDroppingTransformer.ColumnOptions[] dropSlotsColumns;
(string outputColumnName, string inputColumnName)[] copyColumnsPairs;
/// <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));
}
}