keystoneml.nodes

stats

package stats

Visibility
  1. Public
  2. All

Type Members

  1. case class ColumnSampler(numSamplesPerMatrix: Int) extends Transformer[DenseMatrix[Float], DenseMatrix[Float]] with Product with Serializable

    Given a collection of Dense Matrices, this will generate a sample of

  2. class CosineRandomFeatures extends Transformer[DenseVector[Double], DenseVector[Double]]

    Transformer that extracts random cosine features from a feature vector

  3. case class LinearRectifier(maxVal: Double = 0.0, alpha: Double = 0.0) extends Transformer[DenseVector[Double], DenseVector[Double]] with Product with Serializable

    This transformer applies a Linear Rectifier, an activation function defined as: f(x) = max({@param maxVal}, x - {@param alpha})

  4. case class PaddedFFT() extends Transformer[DenseVector[Double], DenseVector[Double]] with Product with Serializable

    This transformer pads input vectors to the nearest power of two, then returns the real values of the first half of the fourier transform on the padded vectors.

  5. case class RandomSignNode(signs: DenseVector[Double]) extends Transformer[DenseVector[Double], DenseVector[Double]] with Product with Serializable

    A node that takes in DenseVector[Double] and randomly flips the sign of some of the elements

  6. class Sampler[T] extends FunctionNode[RDD[T], Array[T]]

    Takes a sample of an input RDD of size size.

  7. class StandardScaler extends Estimator[DenseVector[Double], DenseVector[Double]]

    Standardizes features by removing the mean and scaling to unit std using column summary statistics on the samples in the training set.

  8. class StandardScalerModel extends Transformer[DenseVector[Double], DenseVector[Double]]

    Represents a StandardScaler model that can transform dense vectors.

  9. case class TermFrequency[T](fun: (Double) ⇒ Double = ...) extends Transformer[Seq[T], Seq[(T, Double)]] with Product with Serializable

    Transformer that maps a Seq[Any] of objects to a Seq[(Any, Double)] of (unique object, weighting_scheme(tf)), where tf is the number of times the unique object appeared in the original Seq[Any], and the weighting_scheme is a lambda of Double => Double that defaults to the identity function.

Value Members

  1. object BatchSignedHellingerMapper extends Transformer[DenseMatrix[Float], DenseMatrix[Float]]

  2. object CosineRandomFeatures extends Serializable

    Companion Object to generate random cosine features from various distributions

  3. object NormalizeRows extends Transformer[DenseVector[Double], DenseVector[Double]]

    Divides each row by the max of its two-norm and 2.

  4. object RandomSignNode extends Serializable

  5. object SignedHellingerMapper extends Transformer[DenseVector[Double], DenseVector[Double]]

    Apply power normalization: z <- sign(z)|z|^{\rho} with \rho = \frac{1}{2} This a "signed square root"

Ungrouped