Namespace containing data loading and saving, data schema definitions, and model training metrics components.
Evaluation results for anomaly detection(unsupervised learning algorithm).
Base class for the ISingleFeaturePredictionTransformer<TModel> working on anomaly detection tasks.
Evaluation results for binary classifiers, excluding probabilistic metrics.
This class represents one data point on Precision-Recall curve for binary classification.
Base class for the ISingleFeaturePredictionTransformer<TModel> working on binary classification tasks.
Evaluation results for binary classifiers, including probabilistic metrics.
The metrics generated after evaluating the clustering predictions.
Base class for the ISingleFeaturePredictionTransformer<TModel> working on clustering tasks.
Allows a member to specify IDataView column name directly, as opposed to the default behavior of using the member name as the column name.
This class represents a data loader that applies a transformer chain after loading. It also has methods to save itself to a repository.
An estimator class for composite data loader. It can be used to build a 'trainable smart data loader', although this pattern is not very common.
Represents the confusion matrix of the classification results.
This class represents an eager 'preview' of a IDataView.
This is the abstract base class for all types in the IDataView type system.
DataViewTypeAttribute should be used to decorated class properties and fields, if that class' instances will be loaded as ML.NET IDataView. The function Register() will be called to register a DataViewType for a Type with its Attributes. Whenever a value typed to the registered Type and its Attributes, that value's type (i.e., a Type) in IDataView would be the associated DataViewType.
A singleton class for managing the map between ML.NET DataViewType and C# Type. To support custom column type in IDataView, the column's underlying type (e.g., a C# class's type) should be registered with a class derived from DataViewType.
Represents a chain (potentially empty) of estimators that end with a
Wraps an IFileHandle as an IMultiStreamSource.
This type is for data representing some enumerated value. This is an enumeration over a defined, known cardinality set, as expressed through Count. The underlying .NET type is one of the unsigned integer types. Most commonly this will be UInt32, but could alternately be Byte, UInt16, or UInt64. Despite this, the information is not inherently numeric, so, typically, arithmetic is not meaningful. For example, in multi-class classification, the label is typically a class number which is naturally a KeyDataViewType.
Note that for data of this type, a value of 0, being the default value of the representation type, indicates
the missing value since it would not be sensible for the default value to correspond to any one particular specific
value of the set. The first non-missing value for the enumeration of the set is always
For instance, if you had a key value with a Count of 3, then the UInt32 value
Note that in usage and structure, this is quite close in intended usage and structure to so-called "factor variables" in R.
Allow member to be marked as a KeyDataViewType.
The MetricsStatistics class computes summary statistics over multiple observations of a metric.
Evaluation results for multi-class classification trainers.
Base class for the ISingleFeaturePredictionTransformer<TModel> working on multi-class classification tasks.
Wraps a potentially compound path as an IMultiStreamSource.
The standard number type. This class is not directly instantiable. All allowed instances of this type are singletons, and are accessible as static properties on this class.
Base class for transformer which operates on pairs input and output columns.
Base class for transformers with no feature column, or more than one feature columns.
The abstract base class for all primitive types. Values of these types can be freely copied without concern for ownership, mutation, or disposing.
Evaluation results for rankers.
Base class for the ISingleFeaturePredictionTransformer<TModel> working on ranking tasks.
Evaluation results regression algorithms (supervised learning algorithm).
Base class for the ISingleFeaturePredictionTransformer<TModel> working on regression tasks.
Base class for transformer which produce new columns, but doesn't affect existing ones.
Extension methods to facilitate easy consumption of popular contents of Annotations.
This class defines a schema of a typed data view.
One column of the data view.
A simple disk-based file handle.
The base class for all the transformers implementing the ISingleFeaturePredictionTransformer<TModel>. Those are all the transformers that work with one feature column.
The abstract base class for all non-primitive types.
Loads a text file into an IDataView. Supports basic mapping from input columns to IDataView columns.
Describes how an input column should be mapped to an IDataView column.
The settings for TextLoader
Specifies the range of indices of input columns that should be mapped to an output column.
A chain of transformers (possibly empty) that end with a
Concrete implementations still have to provide the schema propagation mechanism, since there is no easy way to infer it from the transformer.
Various methods for creating VBufferEditor<T> instances.
Allows a member to be marked as a VectorDataViewType, primarily allowing one to set the dimensionality of the resulting array.
A structure serving as the identifier of a row of IDataView. For datasets with millions of records, those IDs need to be unique, therefore the need for such a large structure to hold the values. Those Ids are derived from other Ids of the previous components of the pipelines, and dividing the structure in two: high order and low order of bits, and reduces the changes of those collisions even further.
A buffer that supports both dense and sparse representations. This is the representation type for all VectorDataViewType instances. The explicitly defined values of this vector are exposed through GetValues() and, if not dense, GetIndices().
A file handle.
An interface for exposing some number of items that can be opened for reading.
This interface maps an input DataViewRow to an output DataViewRow. Typically, the output contains both the input columns and new columns added by the implementing class, although some implementations may return a subset of the input columns. This interface is similar to Microsoft.ML.Data.ISchemaBoundRowMapper, except it does not have any input role mappings, so to rebind, the same input column names must be used. Implementations of this interface are typically created over defined input DataViewSchema.
Specifies a simple data type.
This enum allows for 'tagging' the estimators (and subsequently transformers) in the chain to be used 'only for training', 'for training and evaluation' etc. Most notable example is, transformations over the label column should not be used for scoring, so the scope should be Training or TrainTest.