TechnologiesMachine Learning & AI

CreateMLComponents

iOSmacOStvOSwatchOSvisionOS

CreateMLComponents provides composable building blocks for building, training, and evaluating custom machine learning models and feature pipelines on iOS, macOS, tvOS, watchOS, and visionOS. You assemble pipelines from transformers and estimators such as BoostedTreeClassifier and BoostedTreeRegressor, combining preprocessing steps like CategoricalImputer, ColumnSelector, DateFeatureExtractor, and AudioFeaturePrint into composed stages with types like ComposedTransformer and ComposedTabularTransformer. Annotated data structures including AnnotatedFeature, AnnotatedBatch, and AnnotatedFiles carry features and labels through training, while augmentation components such as Augmenter, ApplyRandomly, and ChooseRandomly expand training data. You evaluate results using protocols and metrics like Classifier and ClassificationMetrics, and produce predictions with Classification and ClassificationDistribution.

Pipeline Building Blocks 18

The core protocols that define transformers and estimators you compose into pipelines.

  • Pr
    TransformeriOS 16.0+
    A transformer that takes an input and produces an output.
  • Pr
    EstimatoriOS 16.0+
    An estimator that creates a transformer by fitting to a data set.
  • Pr
    TabularTransformeriOS 16.0+
    A tabular transformer that transforms a data frame.
  • Pr
    TabularEstimatoriOS 16.0+
    A tabular estimator that creates a transformer by fitting to a data set in a data frame.
  • Pr
    TemporalTransformeriOS 16.0+
    A transformer that takes an asynchronous input sequence of temporal features and produces an asynchronous output sequence.
  • Pr
    TemporalEstimator
    An estimator that creates a transformer by fitting to a sequence of temporal features.
  • Pr
    SupervisedEstimatoriOS 16.0+
    An estimator that creates a transformer by fitting to a data set.
  • Pr
    SupervisedTabularEstimatoriOS 16.0+
    A tabular estimator that creates a transformer by fitting to a data set in a data frame.
  • Pr
    SupervisedTemporalEstimator
    An estimator that creates a transformer by fitting to a sequence of annotated temporal features.
  • Pr
    UpdatableEstimatoriOS 16.0+
    An estimator that can be incrementally updated.
  • Pr
    UpdatableSupervisedEstimatoriOS 16.0+
    A supervised estimator that can be incrementally updated.
  • Pr
    UpdatableSupervisedTabularEstimatoriOS 16.0+
    A supervised tabular estimator that can be incrementally updated.
  • Pr
    UpdatableSupervisedTemporalEstimator
    A supervised temporal estimator that can be incrementally updated.
  • Pr
    UpdatableTabularEstimatoriOS 16.0+
    A tabular estimator that can be incrementally updated.
  • Pr
    UpdatableTemporalEstimator
    A temporal estimator that can be incrementally updated.
  • Pr
    EstimatorEncoderiOS 16.0+
    A type that can encode values into a model representation.
  • Pr
    EstimatorDecoderiOS 16.0+
    A type that can decode values from a model representation.
  • Pr
    RandomTransformeriOS 17.0+
    A transformer that takes an input and a random number generator and produces a randomized output.

Composing Pipelines 3

Transformers that chain preprocessing and modeling stages into composed pipelines.

  • St
    ComposedTransformeriOS 16.0+
    A transformer that composes two transformers by applying them one after the other.
  • St
    ComposedTabularTransformeriOS 16.0+
    A transformer that composes two tabular transformers by applying them one after the other.
  • St
    ComposedTemporalTransformeriOS 16.0+
    A temporal transformer that composes two temporal transformers by applying them one after the other.

Adapting Components 21

Adaptors that bridge between transformer, estimator, supervised, and temporal component roles.

  • St
    EstimatorToSupervisedAdaptoriOS 16.0+
    An adaptor that exposes an estimator as a supervised estimator.
  • St
    EstimatorToTemporalAdaptor
    A temporal estimator wrapping an estimator.
  • St
    SupervisedEstimatorToTemporalAdaptor
    A supervised temporal estimator wrapping a supervised estimator.
  • St
    TabularEstimatorToSupervisedAdaptoriOS 16.0+
    An adaptor that exposes a tabular estimator as a tabular supervised estimator.
  • St
    TabularTransformerToEstimatorAdaptoriOS 16.0+
    A tabular estimator that always returns a predefined tabular transformer.
  • St
    TabularTransformerToUpdatableEstimatorAdaptoriOS 16.0+
    An updatable tabular estimator that always returns a predefined transformer.
  • St
    TemporalAdaptoriOS 18.0+
    A temporal transformer that applies a regular transformer to each value of a temporal sequence.
  • St
    TemporalEstimatorToSupervisedAdaptor
    An adaptor that exposes a temporal estimator as a supervised temporal estimator.
  • St
    TemporalTransformerToEstimatorAdaptor
    A temporal estimator that always returns a predefined temporal transformer.
  • St
    TemporalTransformerToUpdatableEstimatorAdaptor
    A temporal estimator that always returns a predefined temporal transformer.
  • St
    TransformerToEstimatorAdaptoriOS 16.0+
    An estimator that always returns a predefined transformer.
  • St
    TransformerToTemporalAdaptor
    A temporal transformer that applies a regular transformer to each value of a temporal sequence.
  • St
    TransformerToUpdatableEstimatorAdaptoriOS 16.0+
    An updatable estimator that always returns a predefined transformer.
  • St
    UpdatableEstimatorToSupervisedAdaptoriOS 16.0+
    An adaptor that exposes an updatable estimator as an updatable supervised estimator.
  • St
    UpdatableEstimatorToTemporalAdaptor
    An updatable temporal estimator wrapping an updatable estimator.
  • St
    UpdatableSupervisedEstimatorToTemporalAdaptor
    An updatable supervised temporal estimator wrapping an updatable supervised estimator.
  • St
    UpdatableTabularEstimatorToSupervisedAdaptoriOS 16.0+
    An adaptor that exposes an updatable tabular estimator as an updatable supervised tabular estimator.
  • St
    UpdatableTemporalEstimatorToSupervisedAdaptor
    An adaptor that exposes an updatable temporal estimator as an updatable supervised temporal estimator.
  • St
    MLModelClassifierAdaptoriOS 16.0+
    A transformer that uses a Core ML model as a classifier.
  • St
    MLModelRegressorAdaptoriOS 16.0+
    A transformer that uses a Core ML model as a regressor.
  • St
    MLModelTransformerAdaptoriOS 16.0+
    A transformer that uses a Core ML model.

Classifiers 9

Estimators and models that learn to assign categorical labels.

  • Pr
    ClassifieriOS 16.0+
    An estimator that predicts classification probabilities.
  • St
    BoostedTreeClassifieriOS 16.0+
    A gradient boosted decision tree classifier.
  • St
    LogisticRegressionClassifieriOS 16.0+
    A logistic regression classifier.
  • St
    LogisticRegressionClassifierModeliOS 16.0+
    A trained logistic regression classifier model.
  • St
    FullyConnectedNetworkClassifieriOS 16.0+
    A classifier that uses a fully connected network.
  • St
    FullyConnectedNetworkClassifierModeliOS 16.0+
    A classifier model that uses a fully connected network.
  • St
    FullyConnectedNetworkMultiLabelClassifieriOS 17.0+
    A classifier that uses a multi-label fully-connected network.
  • St
    FullyConnectedNetworkMultiLabelClassifierModeliOS 17.0+
    A multi-label classifier model that uses a fully-connected network.
  • St
    TreeClassifierModeliOS 16.0+
    A trained tree classifier model.

Regressors 8

Estimators and models that predict continuous numeric values.

  • Pr
    RegressoriOS 16.0+
    A transformer that predicts a float value.
  • St
    BoostedTreeRegressoriOS 16.0+
    A gradient boosted decision tree regressor.
  • St
    LinearRegressoriOS 16.0+
    A linear regressor.
  • St
    LinearRegressorModeliOS 16.0+
    A trained linear regressor model.
  • St
    MultivariateLinearRegressoriOS 18.0+
    A multivariate linear regressor.
  • St
    FullyConnectedNetworkRegressoriOS 16.0+
    A regressor that uses a fully connected network.
  • St
    FullyConnectedNetworkRegressorModeliOS 16.0+
    A regressor model that uses a fully connected network.
  • St
    TreeRegressorModeliOS 16.0+
    A trained tree regressor model.

Time Series & Temporal Models 17

Estimators and sequence types for forecasting and classifying temporal data.

  • St
    LinearTimeSeriesForecasteriOS 18.0+
    A time-series forecasting estimator.
  • St
    LinearTimeSeriesForecasterConfigurationiOS 18.0+
    The configuration for a linear time-series forecaster.
  • St
    TimeSeriesClassifier
    An estimator that trains a model to classify time series sequences.
  • St
    TimeSeriesClassifierConfiguration
    The configuration for a time-series classifier.
  • St
    TimeSeriesForecasterAnnotatedWindowsiOS 18.0+
    A sequence of forecasting windows on a time series shaped array.
  • St
    TimeSeriesForecasterBatchesiOS 18.0+
    A sequence of forecaster batches on a time series shaped array.
  • Pr
    SupervisedTemporalEstimator
    An estimator that creates a transformer by fitting to a sequence of annotated temporal features.
  • Pr
    TemporalSequenceiOS 16.0+
    Async sequence for temporal features.
  • St
    AnyTemporalSequenceiOS 16.0+
    A type-erased temporal sequence.
  • St
    AnyTemporalIteratoriOS 16.0+
    A type-erased async iterator.
  • St
    TemporalFeatureiOS 16.0+
    A temporal feature contains a segment identifier and a feature value.
  • St
    TemporalFileSegmentiOS 16.0+
    A URL and a time range identifying a specific segment of a time-based (temporal) file.
  • St
    TemporalSegmentIdentifieriOS 16.0+
    Uniquely identifiers a segment of a temporal sequence.
  • St
    SlidingWindowsiOS 18.0+
    A sequence of windows on a time series shaped array.
  • St
    SlidingWindowTransformeriOS 16.0+
    A temporal transformer that groups input elements.
  • St
    PreprocessedFeatureSequenceiOS 16.0+
    An asynchronous sequence of eagerly stored temporal features.
  • St
    DataFrameTemporalAnnotationParametersiOS 16.0+
    Annotation parameters for the dataframe containing temporal annotations.

Model Configuration 5

Value types that configure estimator training and capture model metadata.

  • St
    BoostedTreeConfigurationiOS 16.0+
    A boosted tree configuration.
  • St
    FullyConnectedNetworkConfigurationiOS 16.0+
    A fully connected network configuration.
  • St
    MultivariateLinearRegressorConfigurationiOS 18.0+
    A linear regressor configuration.
  • St
    ModelMetadataiOS 17.0+
    User info keys that specify useful information about a model.
  • St
    MetricsKeyiOS 16.0+
    A key that uniquely identifies a metric.

Tabular Preprocessing 19

Transformers that select, impute, scale, encode, and combine tabular feature columns.

  • St
    ColumnSelectoriOS 16.0+
    An operation that applies an estimator to a selection of columns.
  • St
    ColumnSelectorTransformeriOS 16.0+
    A transformer that applies a base transformer to specific columns in a data frame.
  • St
    ColumnConcatenatoriOS 16.0+
    A transformer that concatenates every numerical column in a dataframe into to a shaped array for each row.
  • En
    ColumnSelectioniOS 16.0+
    A selection of columns from a data frame.
  • St
    CategoricalImputeriOS 16.0+
    An estimator that replaces missing values in the categorical input.
  • St
    NumericImputeriOS 16.0+
    An estimator that replaces missing values in the numeric input.
  • St
    ImputeTransformeriOS 16.0+
    A transformer that replaces missing values with a pre-defined value.
  • St
    OneHotEncoderiOS 16.0+
    An estimator that encodes categorical values to an integer array.
  • St
    OrdinalEncoderiOS 16.0+
    An ordinal encoder estimator encodes categorical values to ordinal integer values.
  • St
    MaxAbsScaleriOS 16.0+
    An estimator that scales the input values so that the maximum absolute value is 1.0.
  • St
    MinMaxScaleriOS 16.0+
    An estimator that scales the input values so that they all lie in a closed range.
  • St
    RobustScaleriOS 16.0+
    An estimator that scales the input using statistics that are robust to outliers.
  • St
    StandardScaleriOS 16.0+
    An estimator that standardizes the input by removing the mean and scaling to unit variance.
  • St
    NormalizationScaleriOS 16.0+
    An estimator that normalizes the input values using a normalization strategy.
  • St
    LinearTransformeriOS 16.0+
    A transformer that runs an input through a scale and offset.
  • St
    ReshaperiOS 16.0+
    A transformer that reshapes a shaped array.
  • St
    OptionalUnwrapperiOS 16.0+
    A transformer that unwraps optional elements and throws when encountering missing values.
  • St
    DateFeatureExtractoriOS 18.0+
    A time and date feature extractor.
  • St
    DateFeaturesiOS 18.0+
    A set of date and time features.

Preprocessing Estimators 12

Estimators that pair a learned preprocessing stage with a downstream estimator.

  • St
    PreprocessingEstimatoriOS 16.0+
    An estimator that composes a preprocessing transformer and an estimator.
  • St
    PreprocessingSupervisedEstimatoriOS 16.0+
    A supervised estimator that composes a preprocessing transformer and a supervised estimator.
  • St
    PreprocessingSupervisedTabularEstimatoriOS 17.0+
    A supervised tabular estimator that composes a preprocessing transformer and a supervised tabular estimator.
  • St
    PreprocessingSupervisedTemporalEstimator
    A supervised temporal estimator that composes a preprocessing transformer and a supervised temporal estimator.
  • St
    PreprocessingTabularEstimatoriOS 17.0+
    An estimator that composes a preprocessing transformer and an estimator.
  • St
    PreprocessingTemporalEstimator
    A temporal estimator that composes a preprocessing transformer and a temporal estimator.
  • St
    PreprocessingUpdatableEstimatoriOS 16.0+
    An updatable estimator that composes a preprocessing transformer and an updatable estimator.
  • St
    PreprocessingUpdatableSupervisedEstimatoriOS 16.0+
    An updatable supervised estimator that composes a preprocessing transformer and an updatable supervised estimator.
  • St
    PreprocessingUpdatableSupervisedTabularEstimatoriOS 17.0+
    An updatable supervised estimator that composes a preprocessing transformer and an updatable supervised estimator.
  • St
    PreprocessingUpdatableSupervisedTemporalEstimator
    An updatable supervised temporal estimator that composes a preprocessing transformer and an updatable supervised temporal estimator.
  • St
    PreprocessingUpdatableTabularEstimatoriOS 17.0+
    An updatable estimator that composes a preprocessing transformer and an updatable estimator.
  • St
    PreprocessingUpdatableTemporalEstimator
    An updatable temporal estimator that composes a preprocessing transformer and an updatable temporal estimator.

Image Processing 13

Transformers and extractors that read, transform, and produce features from images.

  • St
    ImageReaderiOS 16.0+
    An image file reader.
  • St
    ImageScaleriOS 16.0+
    An image scaling transformer.
  • St
    ImageCropperiOS 16.0+
    An image crop transformer.
  • St
    ImageRotatoriOS 17.0+
    An image rotating transformer.
  • St
    ImageFlipperiOS 17.0+
    An image flipper transformer.
  • St
    ImageBluriOS 17.0+
    An image blurring transformer.
  • St
    ImageColorTransformeriOS 17.0+
    An image color transformer.
  • St
    ImageExposureAdjusteriOS 17.0+
    An image exposure adjusting transformer.
  • St
    ImageFeaturePrintiOS 16.0+
    ImageFeaturePrint image feature extractor.
  • Pr
    ImageFeatureExtractoriOS 16.0+
    A transformer that takes an image and outputs image features.
  • St
    MLModelImageFeatureExtractoriOS 16.0+
    An image feature extractor provided by an MLModel.
  • St
    RandomImageCropperiOS 17.0+
    Crops an image at a random location.
  • St
    RandomImageNoiseGeneratoriOS 17.0+
    A transformer that adds random noise to an image.

Audio & Video Processing 5

Readers and transformers that ingest and extract features from audio and video.

  • St
    AudioReaderiOS 16.0+
    An audio file reader.
  • St
    AudioConvertingTransformeriOS 16.0+
    A transformer for audio conversion.
  • St
    AudioFeaturePrintiOS 16.0+
    A stream transformer that extracts audio features from audio buffers.
  • St
    DownsampleriOS 16.0+
    A temporal transformer that down samples the input stream.
  • St
    VideoReaderiOS 16.0+
    A video file reader.

Pose & Action Analysis 12

Extractors and types for body and hand pose, object detection, and action recognition.

  • St
    HumanBodyPoseExtractoriOS 16.0+
    The human body pose image feature extractor.
  • St
    HumanHandPoseExtractoriOS 16.0+
    The human hand pose image feature extractor.
  • St
    HumanBodyActionCounteriOS 16.0+
    A human body action repetition counting transformer that takes window of human body poses and produces cumulative human body action repetition counts.
  • St
    HumanBodyActionPeriodPredictoriOS 16.0+
    A human body action period predictor transformer that takes window of poses and produces a window of predictions.
  • St
    PoseiOS 16.0+
    A pose that contains joint keypoints from a person, a hand, or a combination.
  • St
    PoseSelectoriOS 16.0+
    A transformer that selects one pose from an array of poses.
  • En
    PoseSelectionStrategyiOS 16.0+
    Pose selection strategy.
  • St
    JointKeyiOS 16.0+
    A key that uniquely identifies a joint.
  • St
    JointPointiOS 16.0+
    A joint in a pose that contains a location and scoring information.
  • St
    JointsSelectoriOS 16.0+
    Joints selector from a pose.
  • St
    DetectedObjectiOS 17.0+
    An item in a detection result.
  • St
    ObjectDetectionAnnotationiOS 17.0+
    An object detection annotation.

Annotated Data 6

Structures that carry features, labels, and predictions through training and inference.

  • St
    AnnotatedFeatureiOS 16.0+
    An annotated example for fitting a supervised estimator.
  • St
    AnnotatedBatchiOS 18.0+
    A batch of annotated examples for fitting a supervised estimator.
  • St
    AnnotatedFeatureProvideriOS 16.0+
    An adaptor that converts a regular estimator to a tabular estimator by selecting features and annotations from columns.
  • St
    AnnotatedFilesiOS 16.0+
    An annotated files collection.
  • St
    AnnotatedPredictioniOS 17.0+
    An annotated prediction.
  • St
    EventiOS 16.0+
    Maintans the status of the pipeline.

Data Augmentation 7

Components that expand training data by applying random transformations.

  • St
    AugmenteriOS 17.0+
    An augmenter.
  • St
    AugmentationSequenceiOS 17.0+
    An async sequence of augmented elements.
  • St
    UpsampledAugmentationSequenceiOS 17.0+
    An async sequence of augmented elements.
  • St
    ApplyRandomlyiOS 17.0+
    Randomly applies the transformer with the given probability.
  • St
    ApplyEachRandomlyiOS 17.0+
    Applies each transformer randomly given a probability.
  • St
    ChooseRandomlyiOS 17.0+
    Apply single transformation randomly chosen from a list of transformers.
  • St
    ShuffleRandomlyiOS 17.0+
    Apply transformations in a random order.

Predictions & Metrics 5

Types that represent model predictions and evaluate model performance.

  • St
    ClassificationiOS 16.0+
    An item in a classification result.
  • St
    ClassificationDistributioniOS 16.0+
    A classification distribution that contains a probability for each classification label.
  • St
    ClassificationMetricsiOS 16.0+
    Classification metrics.
  • St
    MultiLabelClassificationMetricsiOS 17.0+
    Multi-label classification metrics.
  • St
    ObjectDetectionMetricsiOS 17.0+
    Metrics for object detection model.

Hyperparameter Optimization 3

Parameter ranges and strategies for tuning model hyperparameters.

  • Cl
    UniformRandomIntegerParameteriOS 17.0+
    Applies the transformer with a randomly generated input parameter.
  • St
    UniformRandomFloatingPointParameteriOS 17.0+
    Applies the transformer with a randomly generated input parameter.
  • En
    OptimizationStrategyiOS 17.0+
    A linear optimization strategy.

Errors 13

Error types reported during data loading, preprocessing, training, and serialization.

  • En
    AudioPreprocessingErroriOS 16.0+
    Audio preprocessing errors.
  • En
    AudioReaderErroriOS 16.0+
    Audio reader errors.
  • En
    VideoReaderErroriOS 16.0+
    Video loader errors.
  • En
    CompatibilityErroriOS 17.0+
    A compatibility error.
  • En
    ConcatenationErroriOS 16.0+
    Errors thrown when concatenating numeric values.
  • En
    DatasetErroriOS 16.0+
    Dataset processing errors.
  • En
    EstimatorEncodingErroriOS 16.0+
    An estimator encoding error.
  • En
    ModelCompatibilityErroriOS 16.0+
    Errors related to CoreML model compatibility.
  • En
    ModelUpdateErroriOS 16.0+
    An updatable model error.
  • En
    OptimizationErroriOS 17.0+
    An optimization error.
  • En
    PipelineDataErroriOS 16.0+
    Errors related to pipeline data affinity problems.
  • En
    TabularPipelineDataErroriOS 16.0+
    Errors related to tabular pipeline data affinity problems.
  • En
    SerializationErroriOS 16.0+
    A serialization error.

Structures 1

  • St
    AugmentationBuilderiOS 17.0+
    A series of augmentations.

Type Aliases 1

  • Ty
    EventHandler
    A closure to handle processing events.

Extends 3

DataFrameSequenceLazySequence
← Machine Learning & AI