RandomizedPcaTrainer Class


The IEstimator<TTransformer> for training an approximate PCA using Randomized SVD algorithm.

public sealed class RandomizedPcaTrainer : Microsoft.ML.Trainers.TrainerEstimatorBase<Microsoft.ML.Data.AnomalyPredictionTransformer<Microsoft.ML.Trainers.PcaModelParameters>,Microsoft.ML.Trainers.PcaModelParameters>
type RandomizedPcaTrainer = class
    inherit TrainerEstimatorBase<AnomalyPredictionTransformer<PcaModelParameters>, PcaModelParameters>
Public NotInheritable Class RandomizedPcaTrainer
Inherits TrainerEstimatorBase(Of AnomalyPredictionTransformer(Of PcaModelParameters), PcaModelParameters)


To create this trainer, use RandomizedPca or RandomizedPca(Options).

Input and Output Columns

The input features column data must be a known-sized vector of Single. This trainer outputs the following columns:

Output Column Name Column Type Description
Score Single The non-negative, unbounded score that was calculated by the anomaly detection model.
PredictedLabel Boolean The predicted label, based on the threshold. A score higher than the threshold maps to true and a score lower than the threshold maps to false. The default threshold is 0.5.Use <xref:AnomalyDetectionCatalog.ChangeModelThreshold> to change the default value.

Trainer Characteristics

Machine learning task Anomaly Detection
Is normalization required? Yes
Is caching required? No
Required NuGet in addition to Microsoft.ML None

Training Algorithm Details

This trainer uses the top eigenvectors to approximate the subspace containing the normal class. For each new instance, it computes the norm difference between the raw feature vector and the projected feature on that subspace. If the error is close to 0, the instance is considered normal (non-anomaly).

More specifically, this trainer trains an approximate PCA using a randomized method for computing the singular value decomposition (SVD) of the matrix whose rows are the input vectors. The model generated by this trainer contains three parameters:

  • A projection matrix $U$
  • The mean vector in the original feature space $m$
  • The mean vector in the projected feature space $p$

For an input feature vector $x$, the anomaly score is computed by comparing the $L_2$ norm of the original input vector, and the $L_2$ norm of the projected vector: $\sqrt{\left(|x-m|_2^2 - |Ux-p|_2^2\right)|x-m|_2^2}$.

The method is described here.

Note that the algorithm can be made into Kernel PCA by applying the ApproximatedKernelTransformer to the data before passing it to the trainer.

Check the See Also section for links to usage examples.



The feature column that the trainer expects.

(Inherited from TrainerEstimatorBase<TTransformer,TModel>)

The label column that the trainer expects. Can be null, which indicates that label is not used for training.

(Inherited from TrainerEstimatorBase<TTransformer,TModel>)

The weight column that the trainer expects. Can be null, which indicates that weight is not used for training.

(Inherited from TrainerEstimatorBase<TTransformer,TModel>)





Trains and returns a ITransformer.

(Inherited from TrainerEstimatorBase<TTransformer,TModel>)
GetOutputSchema(SchemaShape) (Inherited from TrainerEstimatorBase<TTransformer,TModel>)

Extension Methods

WithOnFitDelegate<TTransformer>(IEstimator<TTransformer>, Action<TTransformer>)

Given an estimator, return a wrapping object that will call a delegate once Fit(IDataView) is called. It is often important for an estimator to return information about what was fit, which is why the Fit(IDataView) method returns a specifically typed object, rather than just a general ITransformer. However, at the same time, IEstimator<TTransformer> are often formed into pipelines with many objects, so we may need to build a chain of estimators via EstimatorChain<TLastTransformer> where the estimator for which we want to get the transformer is buried somewhere in this chain. For that scenario, we can through this method attach a delegate that will be called once fit is called.

Applies to

See also