File-scoped namespaces in files under `Prediction` (`Microsoft.ML.Core`) (#6792)
Co-authored-by: Lehonti Ramos <john@doe>
This commit is contained in:
Lehonti Ramos
GitHub
Родитель
e6a88c440b
Коммит
ccf34e370b
|
|
@ -2,62 +2,61 @@
|
|||
// The .NET Foundation licenses this file to you under the MIT license.
|
||||
// See the LICENSE file in the project root for more information.
|
||||
|
||||
namespace Microsoft.ML
|
||||
namespace Microsoft.ML;
|
||||
|
||||
/// <summary>
|
||||
/// Type of prediction task. Note that this is a legacy structure and usage of this should generally be
|
||||
/// discouraged in future projects. Its presence suggests that there are privileged and supported
|
||||
/// tasks, and anything outside of this is unsupported. This runs rather contrary to the idea of this
|
||||
/// being an expandable framework, and it is inappropriately limiting. For legacy pipelines based on
|
||||
/// <see cref="ITrainer"/> and <see cref="IPredictor"/> it is still useful, but for things based on
|
||||
/// the <see cref="IEstimator{TTransformer}"/> idiom, it is inappropriate.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal enum PredictionKind
|
||||
{
|
||||
Unknown = 0,
|
||||
Custom = 1,
|
||||
|
||||
BinaryClassification = 2,
|
||||
MulticlassClassification = 3,
|
||||
Regression = 4,
|
||||
MultiOutputRegression = 5,
|
||||
Ranking = 6,
|
||||
Recommendation = 7,
|
||||
AnomalyDetection = 8,
|
||||
Clustering = 9,
|
||||
SequenceClassification = 10,
|
||||
|
||||
// More to be added later.
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Weakly typed version of IPredictor.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IPredictor
|
||||
{
|
||||
/// <summary>
|
||||
/// Type of prediction task. Note that this is a legacy structure and usage of this should generally be
|
||||
/// discouraged in future projects. Its presence suggests that there are privileged and supported
|
||||
/// tasks, and anything outside of this is unsupported. This runs rather contrary to the idea of this
|
||||
/// being an expandable framework, and it is inappropriately limiting. For legacy pipelines based on
|
||||
/// <see cref="ITrainer"/> and <see cref="IPredictor"/> it is still useful, but for things based on
|
||||
/// the <see cref="IEstimator{TTransformer}"/> idiom, it is inappropriate.
|
||||
/// Return the type of prediction task.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal enum PredictionKind
|
||||
{
|
||||
Unknown = 0,
|
||||
Custom = 1,
|
||||
|
||||
BinaryClassification = 2,
|
||||
MulticlassClassification = 3,
|
||||
Regression = 4,
|
||||
MultiOutputRegression = 5,
|
||||
Ranking = 6,
|
||||
Recommendation = 7,
|
||||
AnomalyDetection = 8,
|
||||
Clustering = 9,
|
||||
SequenceClassification = 10,
|
||||
|
||||
// More to be added later.
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Weakly typed version of IPredictor.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IPredictor
|
||||
{
|
||||
/// <summary>
|
||||
/// Return the type of prediction task.
|
||||
/// </summary>
|
||||
PredictionKind PredictionKind { get; }
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// A predictor the produces values of the indicated type.
|
||||
/// REVIEW: Determine whether this is just a temporary shim or long term solution.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IPredictorProducing<out TResult> : IPredictor
|
||||
{
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// A predictor that produces values and distributions of the indicated types.
|
||||
/// Note that from a public API perspective this is bad.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IDistPredictorProducing<out TResult, out TResultDistribution> : IPredictorProducing<TResult>
|
||||
{
|
||||
}
|
||||
PredictionKind PredictionKind { get; }
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// A predictor the produces values of the indicated type.
|
||||
/// REVIEW: Determine whether this is just a temporary shim or long term solution.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IPredictorProducing<out TResult> : IPredictor
|
||||
{
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// A predictor that produces values and distributions of the indicated types.
|
||||
/// Note that from a public API perspective this is bad.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface IDistPredictorProducing<out TResult, out TResultDistribution> : IPredictorProducing<TResult>
|
||||
{
|
||||
}
|
||||
|
|
|
|||
|
|
@ -4,104 +4,103 @@
|
|||
|
||||
using Microsoft.ML.Data;
|
||||
|
||||
namespace Microsoft.ML
|
||||
namespace Microsoft.ML;
|
||||
|
||||
// REVIEW: Would be nice if the registration under SignatureTrainer were automatic
|
||||
// given registration for one of the "sub-class" signatures.
|
||||
|
||||
/// <summary>
|
||||
/// Loadable class signatures for trainers. Typically each trainer should register with
|
||||
/// both SignatureTrainer and SignatureXxxTrainer where Xxx is the prediction kind.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal delegate void SignatureTrainer();
|
||||
|
||||
[BestFriend]
|
||||
internal delegate void SignatureBinaryClassifierTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMulticlassClassifierTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureRegressorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMultiOutputRegressorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureRankerTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureAnomalyDetectorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureClusteringTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureSequenceTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMatrixRecommendingTrainer();
|
||||
|
||||
/// <summary>
|
||||
/// The base interface for a trainers. Implementors should not implement this interface directly,
|
||||
/// but rather implement the more specific <see cref="ITrainer{TPredictor}"/>.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITrainer
|
||||
{
|
||||
// REVIEW: Would be nice if the registration under SignatureTrainer were automatic
|
||||
// given registration for one of the "sub-class" signatures.
|
||||
/// <summary>
|
||||
/// Auxiliary information about the trainer in terms of its capabilities
|
||||
/// and requirements.
|
||||
/// </summary>
|
||||
TrainerInfo Info { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Loadable class signatures for trainers. Typically each trainer should register with
|
||||
/// both SignatureTrainer and SignatureXxxTrainer where Xxx is the prediction kind.
|
||||
/// Return the type of prediction task for the produced predictor.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal delegate void SignatureTrainer();
|
||||
|
||||
[BestFriend]
|
||||
internal delegate void SignatureBinaryClassifierTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMulticlassClassifierTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureRegressorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMultiOutputRegressorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureRankerTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureAnomalyDetectorTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureClusteringTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureSequenceTrainer();
|
||||
[BestFriend]
|
||||
internal delegate void SignatureMatrixRecommendingTrainer();
|
||||
PredictionKind PredictionKind { get; }
|
||||
|
||||
/// <summary>
|
||||
/// The base interface for a trainers. Implementors should not implement this interface directly,
|
||||
/// but rather implement the more specific <see cref="ITrainer{TPredictor}"/>.
|
||||
/// Trains a predictor.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITrainer
|
||||
{
|
||||
/// <summary>
|
||||
/// Auxiliary information about the trainer in terms of its capabilities
|
||||
/// and requirements.
|
||||
/// </summary>
|
||||
TrainerInfo Info { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Return the type of prediction task for the produced predictor.
|
||||
/// </summary>
|
||||
PredictionKind PredictionKind { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Trains a predictor.
|
||||
/// </summary>
|
||||
/// <param name="context">A context containing at least the training data</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
/// <seealso cref="ITrainer{TPredictor}.Train(TrainContext)"/>
|
||||
IPredictor Train(TrainContext context);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Strongly typed generic interface for a trainer. A trainer object takes training data
|
||||
/// and produces a predictor.
|
||||
/// </summary>
|
||||
/// <typeparam name="TPredictor"> Type of predictor produced</typeparam>
|
||||
[BestFriend]
|
||||
internal interface ITrainer<out TPredictor> : ITrainer
|
||||
where TPredictor : IPredictor
|
||||
{
|
||||
/// <summary>
|
||||
/// Trains a predictor.
|
||||
/// </summary>
|
||||
/// <param name="context">A context containing at least the training data</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
new TPredictor Train(TrainContext context);
|
||||
}
|
||||
|
||||
[BestFriend]
|
||||
internal static class TrainerExtensions
|
||||
{
|
||||
/// <summary>
|
||||
/// Convenience train extension for the case where one has only a training set with no auxiliary information.
|
||||
/// Equivalent to calling <see cref="ITrainer.Train(TrainContext)"/>
|
||||
/// on a <see cref="TrainContext"/> constructed with <paramref name="trainData"/>.
|
||||
/// </summary>
|
||||
/// <param name="trainer">The trainer</param>
|
||||
/// <param name="trainData">The training data.</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
public static IPredictor Train(this ITrainer trainer, RoleMappedData trainData)
|
||||
=> trainer.Train(new TrainContext(trainData));
|
||||
|
||||
/// <summary>
|
||||
/// Convenience train extension for the case where one has only a training set with no auxiliary information.
|
||||
/// Equivalent to calling <see cref="ITrainer{TPredictor}.Train(TrainContext)"/>
|
||||
/// on a <see cref="TrainContext"/> constructed with <paramref name="trainData"/>.
|
||||
/// </summary>
|
||||
/// <param name="trainer">The trainer</param>
|
||||
/// <param name="trainData">The training data.</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
public static TPredictor Train<TPredictor>(this ITrainer<TPredictor> trainer, RoleMappedData trainData) where TPredictor : IPredictor
|
||||
=> trainer.Train(new TrainContext(trainData));
|
||||
}
|
||||
/// <param name="context">A context containing at least the training data</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
/// <seealso cref="ITrainer{TPredictor}.Train(TrainContext)"/>
|
||||
IPredictor Train(TrainContext context);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Strongly typed generic interface for a trainer. A trainer object takes training data
|
||||
/// and produces a predictor.
|
||||
/// </summary>
|
||||
/// <typeparam name="TPredictor"> Type of predictor produced</typeparam>
|
||||
[BestFriend]
|
||||
internal interface ITrainer<out TPredictor> : ITrainer
|
||||
where TPredictor : IPredictor
|
||||
{
|
||||
/// <summary>
|
||||
/// Trains a predictor.
|
||||
/// </summary>
|
||||
/// <param name="context">A context containing at least the training data</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
new TPredictor Train(TrainContext context);
|
||||
}
|
||||
|
||||
[BestFriend]
|
||||
internal static class TrainerExtensions
|
||||
{
|
||||
/// <summary>
|
||||
/// Convenience train extension for the case where one has only a training set with no auxiliary information.
|
||||
/// Equivalent to calling <see cref="ITrainer.Train(TrainContext)"/>
|
||||
/// on a <see cref="TrainContext"/> constructed with <paramref name="trainData"/>.
|
||||
/// </summary>
|
||||
/// <param name="trainer">The trainer</param>
|
||||
/// <param name="trainData">The training data.</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
public static IPredictor Train(this ITrainer trainer, RoleMappedData trainData)
|
||||
=> trainer.Train(new TrainContext(trainData));
|
||||
|
||||
/// <summary>
|
||||
/// Convenience train extension for the case where one has only a training set with no auxiliary information.
|
||||
/// Equivalent to calling <see cref="ITrainer{TPredictor}.Train(TrainContext)"/>
|
||||
/// on a <see cref="TrainContext"/> constructed with <paramref name="trainData"/>.
|
||||
/// </summary>
|
||||
/// <param name="trainer">The trainer</param>
|
||||
/// <param name="trainData">The training data.</param>
|
||||
/// <returns>The trained predictor</returns>
|
||||
public static TPredictor Train<TPredictor>(this ITrainer<TPredictor> trainer, RoleMappedData trainData) where TPredictor : IPredictor
|
||||
=> trainer.Train(new TrainContext(trainData));
|
||||
}
|
||||
|
|
|
|||
|
|
@ -4,139 +4,138 @@
|
|||
|
||||
using System.Collections.Generic;
|
||||
|
||||
namespace Microsoft.ML.TreePredictor
|
||||
namespace Microsoft.ML.TreePredictor;
|
||||
|
||||
// The interfaces contained herein are meant to allow tree visualizer to run without an explicit dependency
|
||||
// on FastTree, so as to allow it greater generality. These should probably be moved somewhere else, but where?
|
||||
// FastTree itself is not a good candidate since their entire purpose was to avoid tying the tree visualizer
|
||||
// to FastTree itself. They are semi-tolerable though as a set of internal types here.
|
||||
|
||||
/// <summary>
|
||||
/// Predictor that has ensemble tree structures and returns collection of trees.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITreeEnsemble
|
||||
{
|
||||
// The interfaces contained herein are meant to allow tree visualizer to run without an explicit dependency
|
||||
// on FastTree, so as to allow it greater generality. These should probably be moved somewhere else, but where?
|
||||
// FastTree itself is not a good candidate since their entire purpose was to avoid tying the tree visualizer
|
||||
// to FastTree itself. They are semi-tolerable though as a set of internal types here.
|
||||
/// <summary>
|
||||
/// Returns the number of trees in the ensemble.
|
||||
/// </summary>
|
||||
int NumTrees { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Predictor that has ensemble tree structures and returns collection of trees.
|
||||
/// Returns the collection of trees.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITreeEnsemble
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns the number of trees in the ensemble.
|
||||
/// </summary>
|
||||
int NumTrees { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns the collection of trees.
|
||||
/// </summary>
|
||||
/// <returns>Collection of trees</returns>
|
||||
ITree[] GetTrees();
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type of tree used in ensemble of tree based predictors
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITree
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns the array of right(Greater than) child nodes of every interior nodes
|
||||
/// </summary>
|
||||
int[] GtChild { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns the array of left(Leser than or equal to) nodes of every interior nodes
|
||||
/// </summary>
|
||||
int[] LteChild { get; }
|
||||
|
||||
/// <summary>
|
||||
/// returns the number of interior nodes.
|
||||
/// </summary>
|
||||
int NumNodes { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns the number of leaf nodes.
|
||||
/// </summary>
|
||||
int NumLeaves { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns node structure for the given node
|
||||
/// </summary>
|
||||
/// <param name="nodeId">Node id</param>
|
||||
/// <param name="isLeaf">Flag to denote whether the node is leaf or not</param>
|
||||
/// <param name="featureNames">Feature names collection</param>
|
||||
/// <returns>Node structure</returns>
|
||||
INode GetNode(int nodeId, bool isLeaf, IEnumerable<string> featureNames);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type of tree used in ensemble of tree based predictors
|
||||
/// </summary>
|
||||
/// <typeparam name="TFeatures">Type of features container (instance) on which to make predictions</typeparam>
|
||||
[BestFriend]
|
||||
internal interface ITree<TFeatures> : ITree
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns the leaf node for the given instance.
|
||||
/// </summary>
|
||||
/// <param name="features">Type of features container (instance) on which to make predictions</param>
|
||||
/// <returns>node id</returns>
|
||||
/// <typeparamref name="TFeatures">Type of features container (instance) on which to make predictions</typeparamref>
|
||||
int GetLeaf(in TFeatures features);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type to represent the structure of node
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface INode
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns Key value pairs representing the properties of the node.
|
||||
/// </summary>
|
||||
Dictionary<string, object> KeyValues { get; }
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Keys to represent the properties of node.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal static class NodeKeys
|
||||
{
|
||||
/// <summary>
|
||||
/// Name of the the interior node. It is Feature name if it is fasttree. Type is string for default trees.
|
||||
/// </summary>
|
||||
public const string SplitName = "SplitName";
|
||||
|
||||
/// <summary>
|
||||
/// Split gain of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string SplitGain = "SplitGain";
|
||||
|
||||
/// <summary>
|
||||
/// Threshold value of the interior node. Type is string for default trees.
|
||||
/// It is expected that the string has exactly two space separated components.
|
||||
/// i. The first one should be the operator
|
||||
/// ii. The second one should be the actual threshold
|
||||
/// For ex., for a split like f1 <= 10, expected Threshold is "<= 10"
|
||||
/// For a split like color not-in { blue, green }, expected Threshold is "not-in { blue, green }"
|
||||
/// </summary>
|
||||
public const string Threshold = "Threshold";
|
||||
|
||||
/// <summary>
|
||||
/// Gain value (specific to fasttree) of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string GainValue = "GainValue";
|
||||
|
||||
/// <summary>
|
||||
/// Previous leaf value(specific to fasttree) of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string PreviousLeafValue = "PreviousLeafValue";
|
||||
|
||||
/// <summary>
|
||||
/// Leaf value of the leaf node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string LeafValue = "LeafValue";
|
||||
|
||||
/// <summary>
|
||||
/// Extra items that will be displayed in the tool-tip. Type is string for default trees.
|
||||
/// </summary>
|
||||
public const string Extras = "Extras";
|
||||
}
|
||||
/// <returns>Collection of trees</returns>
|
||||
ITree[] GetTrees();
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type of tree used in ensemble of tree based predictors
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface ITree
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns the array of right(Greater than) child nodes of every interior nodes
|
||||
/// </summary>
|
||||
int[] GtChild { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns the array of left(Leser than or equal to) nodes of every interior nodes
|
||||
/// </summary>
|
||||
int[] LteChild { get; }
|
||||
|
||||
/// <summary>
|
||||
/// returns the number of interior nodes.
|
||||
/// </summary>
|
||||
int NumNodes { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns the number of leaf nodes.
|
||||
/// </summary>
|
||||
int NumLeaves { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Returns node structure for the given node
|
||||
/// </summary>
|
||||
/// <param name="nodeId">Node id</param>
|
||||
/// <param name="isLeaf">Flag to denote whether the node is leaf or not</param>
|
||||
/// <param name="featureNames">Feature names collection</param>
|
||||
/// <returns>Node structure</returns>
|
||||
INode GetNode(int nodeId, bool isLeaf, IEnumerable<string> featureNames);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type of tree used in ensemble of tree based predictors
|
||||
/// </summary>
|
||||
/// <typeparam name="TFeatures">Type of features container (instance) on which to make predictions</typeparam>
|
||||
[BestFriend]
|
||||
internal interface ITree<TFeatures> : ITree
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns the leaf node for the given instance.
|
||||
/// </summary>
|
||||
/// <param name="features">Type of features container (instance) on which to make predictions</param>
|
||||
/// <returns>node id</returns>
|
||||
/// <typeparamref name="TFeatures">Type of features container (instance) on which to make predictions</typeparamref>
|
||||
int GetLeaf(in TFeatures features);
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Type to represent the structure of node
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal interface INode
|
||||
{
|
||||
/// <summary>
|
||||
/// Returns Key value pairs representing the properties of the node.
|
||||
/// </summary>
|
||||
Dictionary<string, object> KeyValues { get; }
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Keys to represent the properties of node.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal static class NodeKeys
|
||||
{
|
||||
/// <summary>
|
||||
/// Name of the the interior node. It is Feature name if it is fasttree. Type is string for default trees.
|
||||
/// </summary>
|
||||
public const string SplitName = "SplitName";
|
||||
|
||||
/// <summary>
|
||||
/// Split gain of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string SplitGain = "SplitGain";
|
||||
|
||||
/// <summary>
|
||||
/// Threshold value of the interior node. Type is string for default trees.
|
||||
/// It is expected that the string has exactly two space separated components.
|
||||
/// i. The first one should be the operator
|
||||
/// ii. The second one should be the actual threshold
|
||||
/// For ex., for a split like f1 <= 10, expected Threshold is "<= 10"
|
||||
/// For a split like color not-in { blue, green }, expected Threshold is "not-in { blue, green }"
|
||||
/// </summary>
|
||||
public const string Threshold = "Threshold";
|
||||
|
||||
/// <summary>
|
||||
/// Gain value (specific to fasttree) of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string GainValue = "GainValue";
|
||||
|
||||
/// <summary>
|
||||
/// Previous leaf value(specific to fasttree) of the interior node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string PreviousLeafValue = "PreviousLeafValue";
|
||||
|
||||
/// <summary>
|
||||
/// Leaf value of the leaf node. Type is double for default trees.
|
||||
/// </summary>
|
||||
public const string LeafValue = "LeafValue";
|
||||
|
||||
/// <summary>
|
||||
/// Extra items that will be displayed in the tool-tip. Type is string for default trees.
|
||||
/// </summary>
|
||||
public const string Extras = "Extras";
|
||||
}
|
||||
|
|
|
|||
|
|
@ -5,64 +5,63 @@
|
|||
using Microsoft.ML.Data;
|
||||
using Microsoft.ML.Runtime;
|
||||
|
||||
namespace Microsoft.ML
|
||||
namespace Microsoft.ML;
|
||||
|
||||
/// <summary>
|
||||
/// Holds information relevant to trainers. Instances of this class are meant to be constructed and passed
|
||||
/// into <see cref="ITrainer{TPredictor}.Train(TrainContext)"/> or <see cref="ITrainer.Train(TrainContext)"/>.
|
||||
/// This holds at least a training set, as well as optionally a predictor.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal sealed class TrainContext
|
||||
{
|
||||
/// <summary>
|
||||
/// Holds information relevant to trainers. Instances of this class are meant to be constructed and passed
|
||||
/// into <see cref="ITrainer{TPredictor}.Train(TrainContext)"/> or <see cref="ITrainer.Train(TrainContext)"/>.
|
||||
/// This holds at least a training set, as well as optionally a predictor.
|
||||
/// The training set. Cannot be <c>null</c>.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal sealed class TrainContext
|
||||
public RoleMappedData TrainingSet { get; }
|
||||
|
||||
/// <summary>
|
||||
/// The validation set. Can be <c>null</c>. Note that passing a non-<c>null</c> validation set into
|
||||
/// a trainer that does not support validation sets should not be considered an error condition. It
|
||||
/// should simply be ignored in that case.
|
||||
/// </summary>
|
||||
public RoleMappedData ValidationSet { get; }
|
||||
|
||||
/// <summary>
|
||||
/// The test set, whose uses are very similar to validation set but it should not directly and indirectly
|
||||
/// affect the training process. One major difference between validation set and test test is that validation
|
||||
/// can affect the training process by, for example, early stopping. Note that early stopping is a technique
|
||||
/// which terminates the training process once the scores computed on validation set starts getting worse.
|
||||
/// </summary>
|
||||
public RoleMappedData TestSet { get; }
|
||||
|
||||
/// <summary>
|
||||
/// The initial predictor, for incremental training. Note that if a <see cref="ITrainer"/> implementor
|
||||
/// does not support incremental training, then it can ignore it similarly to how one would ignore
|
||||
/// <see cref="ValidationSet"/>. However, if the trainer does support incremental training and there
|
||||
/// is something wrong with a non-<c>null</c> value of this, then the trainer ought to throw an exception.
|
||||
/// </summary>
|
||||
public IPredictor InitialPredictor { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Constructor, given a training set and optional other arguments.
|
||||
/// </summary>
|
||||
/// <param name="trainingSet">Will set <see cref="TrainingSet"/> to this value. This must be specified</param>
|
||||
/// <param name="validationSet">Will set <see cref="ValidationSet"/> to this value if specified</param>
|
||||
/// <param name="testSet">Will set <see cref="TestSet"/> to this value if specified</param>
|
||||
/// <param name="initialPredictor">Will set <see cref="InitialPredictor"/> to this value if specified</param>
|
||||
public TrainContext(RoleMappedData trainingSet, RoleMappedData validationSet = null, RoleMappedData testSet = null, IPredictor initialPredictor = null)
|
||||
{
|
||||
/// <summary>
|
||||
/// The training set. Cannot be <c>null</c>.
|
||||
/// </summary>
|
||||
public RoleMappedData TrainingSet { get; }
|
||||
Contracts.CheckValue(trainingSet, nameof(trainingSet));
|
||||
Contracts.CheckValueOrNull(validationSet);
|
||||
Contracts.CheckValueOrNull(initialPredictor);
|
||||
|
||||
/// <summary>
|
||||
/// The validation set. Can be <c>null</c>. Note that passing a non-<c>null</c> validation set into
|
||||
/// a trainer that does not support validation sets should not be considered an error condition. It
|
||||
/// should simply be ignored in that case.
|
||||
/// </summary>
|
||||
public RoleMappedData ValidationSet { get; }
|
||||
// REVIEW: Should there be code here to ensure that the role mappings between the two are compatible?
|
||||
// That is, all the role mappings are the same and the columns between them have identical types?
|
||||
|
||||
/// <summary>
|
||||
/// The test set, whose uses are very similar to validation set but it should not directly and indirectly
|
||||
/// affect the training process. One major difference between validation set and test test is that validation
|
||||
/// can affect the training process by, for example, early stopping. Note that early stopping is a technique
|
||||
/// which terminates the training process once the scores computed on validation set starts getting worse.
|
||||
/// </summary>
|
||||
public RoleMappedData TestSet { get; }
|
||||
|
||||
/// <summary>
|
||||
/// The initial predictor, for incremental training. Note that if a <see cref="ITrainer"/> implementor
|
||||
/// does not support incremental training, then it can ignore it similarly to how one would ignore
|
||||
/// <see cref="ValidationSet"/>. However, if the trainer does support incremental training and there
|
||||
/// is something wrong with a non-<c>null</c> value of this, then the trainer ought to throw an exception.
|
||||
/// </summary>
|
||||
public IPredictor InitialPredictor { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Constructor, given a training set and optional other arguments.
|
||||
/// </summary>
|
||||
/// <param name="trainingSet">Will set <see cref="TrainingSet"/> to this value. This must be specified</param>
|
||||
/// <param name="validationSet">Will set <see cref="ValidationSet"/> to this value if specified</param>
|
||||
/// <param name="testSet">Will set <see cref="TestSet"/> to this value if specified</param>
|
||||
/// <param name="initialPredictor">Will set <see cref="InitialPredictor"/> to this value if specified</param>
|
||||
public TrainContext(RoleMappedData trainingSet, RoleMappedData validationSet = null, RoleMappedData testSet = null, IPredictor initialPredictor = null)
|
||||
{
|
||||
Contracts.CheckValue(trainingSet, nameof(trainingSet));
|
||||
Contracts.CheckValueOrNull(validationSet);
|
||||
Contracts.CheckValueOrNull(initialPredictor);
|
||||
|
||||
// REVIEW: Should there be code here to ensure that the role mappings between the two are compatible?
|
||||
// That is, all the role mappings are the same and the columns between them have identical types?
|
||||
|
||||
TrainingSet = trainingSet;
|
||||
ValidationSet = validationSet;
|
||||
TestSet = testSet;
|
||||
InitialPredictor = initialPredictor;
|
||||
}
|
||||
TrainingSet = trainingSet;
|
||||
ValidationSet = validationSet;
|
||||
TestSet = testSet;
|
||||
InitialPredictor = initialPredictor;
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -2,83 +2,82 @@
|
|||
// The .NET Foundation licenses this file to you under the MIT license.
|
||||
// See the LICENSE file in the project root for more information.
|
||||
|
||||
namespace Microsoft.ML
|
||||
namespace Microsoft.ML;
|
||||
|
||||
/// <summary>
|
||||
/// Characteristics of a trainer. Exposed via the Info property of each trainer.
|
||||
/// </summary>
|
||||
public sealed class TrainerInfo
|
||||
{
|
||||
// REVIEW: Ideally trainers should be able to communicate
|
||||
// something about the type of data they are capable of being trained
|
||||
// on, for example, what ColumnKinds they want, how many of each, of what type,
|
||||
// etc. This interface seems like the most natural conduit for that sort
|
||||
// of extra information.
|
||||
|
||||
/// <summary>
|
||||
/// Characteristics of a trainer. Exposed via the Info property of each trainer.
|
||||
/// Whether the trainer needs to see data in normalized form. Only non-parametric trainers will tend to return
|
||||
/// <see langword="false"/> here.
|
||||
/// </summary>
|
||||
public sealed class TrainerInfo
|
||||
public bool NeedNormalization { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer needs calibration to produce probabilities. As a general rule only trainers that produce
|
||||
/// binary classifier predictors that also do not have a natural probabilistic interpretation should have a
|
||||
/// <see langword="true"/> value here.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool NeedCalibration { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether this trainer could benefit from a cached view of the data. Trainers that have few passes over the
|
||||
/// data, or that need to build their own custom data structure over the data, will have a <c>false</c> here.
|
||||
/// </summary>
|
||||
public bool WantCaching { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer supports validation set via <see cref="TrainContext.ValidationSet"/>. Not implementing
|
||||
/// this interface and returning <c>false</c> from this property is an indication the trainer does not support
|
||||
/// that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsValidation { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer can use test set via <see cref="TrainContext.TestSet"/>. Not implementing
|
||||
/// this interface and returning <c>false</c> from this property is an indication the trainer does not support
|
||||
/// that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsTest { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer can support incremental trainers via <see cref="TrainContext.InitialPredictor"/>. Not
|
||||
/// implementing this interface and returning <c>true</c> from this property is an indication the trainer does
|
||||
/// not support that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsIncrementalTraining { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Initializes with the given parameters. The parameters have default values for the most typical values
|
||||
/// for most classical trainers.
|
||||
/// </summary>
|
||||
/// <param name="normalization">The value for the property <see cref="NeedNormalization"/></param>
|
||||
/// <param name="calibration">The value for the property <see cref="NeedCalibration"/></param>
|
||||
/// <param name="caching">The value for the property <see cref="WantCaching"/></param>
|
||||
/// <param name="supportValid">The value for the property <see cref="SupportsValidation"/></param>
|
||||
/// <param name="supportIncrementalTrain">The value for the property <see cref="SupportsIncrementalTraining"/></param>
|
||||
/// <param name="supportTest">The value for the property <see cref="SupportsTest"/></param>
|
||||
[BestFriend]
|
||||
internal TrainerInfo(bool normalization = true, bool calibration = false, bool caching = true,
|
||||
bool supportValid = false, bool supportIncrementalTrain = false, bool supportTest = false)
|
||||
{
|
||||
// REVIEW: Ideally trainers should be able to communicate
|
||||
// something about the type of data they are capable of being trained
|
||||
// on, for example, what ColumnKinds they want, how many of each, of what type,
|
||||
// etc. This interface seems like the most natural conduit for that sort
|
||||
// of extra information.
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer needs to see data in normalized form. Only non-parametric trainers will tend to return
|
||||
/// <see langword="false"/> here.
|
||||
/// </summary>
|
||||
public bool NeedNormalization { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer needs calibration to produce probabilities. As a general rule only trainers that produce
|
||||
/// binary classifier predictors that also do not have a natural probabilistic interpretation should have a
|
||||
/// <see langword="true"/> value here.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool NeedCalibration { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether this trainer could benefit from a cached view of the data. Trainers that have few passes over the
|
||||
/// data, or that need to build their own custom data structure over the data, will have a <c>false</c> here.
|
||||
/// </summary>
|
||||
public bool WantCaching { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer supports validation set via <see cref="TrainContext.ValidationSet"/>. Not implementing
|
||||
/// this interface and returning <c>false</c> from this property is an indication the trainer does not support
|
||||
/// that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsValidation { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer can use test set via <see cref="TrainContext.TestSet"/>. Not implementing
|
||||
/// this interface and returning <c>false</c> from this property is an indication the trainer does not support
|
||||
/// that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsTest { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Whether the trainer can support incremental trainers via <see cref="TrainContext.InitialPredictor"/>. Not
|
||||
/// implementing this interface and returning <c>true</c> from this property is an indication the trainer does
|
||||
/// not support that.
|
||||
/// </summary>
|
||||
[BestFriend]
|
||||
internal bool SupportsIncrementalTraining { get; }
|
||||
|
||||
/// <summary>
|
||||
/// Initializes with the given parameters. The parameters have default values for the most typical values
|
||||
/// for most classical trainers.
|
||||
/// </summary>
|
||||
/// <param name="normalization">The value for the property <see cref="NeedNormalization"/></param>
|
||||
/// <param name="calibration">The value for the property <see cref="NeedCalibration"/></param>
|
||||
/// <param name="caching">The value for the property <see cref="WantCaching"/></param>
|
||||
/// <param name="supportValid">The value for the property <see cref="SupportsValidation"/></param>
|
||||
/// <param name="supportIncrementalTrain">The value for the property <see cref="SupportsIncrementalTraining"/></param>
|
||||
/// <param name="supportTest">The value for the property <see cref="SupportsTest"/></param>
|
||||
[BestFriend]
|
||||
internal TrainerInfo(bool normalization = true, bool calibration = false, bool caching = true,
|
||||
bool supportValid = false, bool supportIncrementalTrain = false, bool supportTest = false)
|
||||
{
|
||||
NeedNormalization = normalization;
|
||||
NeedCalibration = calibration;
|
||||
WantCaching = caching;
|
||||
SupportsValidation = supportValid;
|
||||
SupportsIncrementalTraining = supportIncrementalTrain;
|
||||
SupportsTest = supportTest;
|
||||
}
|
||||
NeedNormalization = normalization;
|
||||
NeedCalibration = calibration;
|
||||
WantCaching = caching;
|
||||
SupportsValidation = supportValid;
|
||||
SupportsIncrementalTraining = supportIncrementalTrain;
|
||||
SupportsTest = supportTest;
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Загрузка…
Ссылка в новой задаче