Package

org.apache.flink.ml

common

Permalink

package common

Visibility
  1. Public
  2. All

Type Members

  1. case class Block[T](index: Int, values: Vector[T]) extends Product with Serializable

    Permalink

    Base class for blocks of elements.

    Base class for blocks of elements.

    TODO: Replace Vector type by Array type once Flink supports generic arrays

  2. case class LabeledVector(label: Double, vector: Vector) extends Serializable with Product

    Permalink

    This class represents a vector with an associated label as it is required for many supervised learning tasks.

    This class represents a vector with an associated label as it is required for many supervised learning tasks.

    label

    Label of the data point

    vector

    Data point

  3. trait Parameter[T] extends AnyRef

    Permalink

    Base trait for parameter keys

    Base trait for parameter keys

    T

    Type of parameter value associated to this parameter key

  4. class ParameterMap extends Serializable

    Permalink

    Map used to store configuration parameters for algorithms.

    Map used to store configuration parameters for algorithms. The parameter values are stored in a Map being identified by a Parameter object. ParameterMaps can be fused. This operation is left associative, meaning that latter ParameterMaps can override parameter values defined in a preceding ParameterMap.

  5. case class WeightVector(weights: Vector, intercept: Double) extends Serializable with Product

    Permalink

    This class represents a weight vector with an intercept, as it is required for many supervised learning tasks

    This class represents a weight vector with an intercept, as it is required for many supervised learning tasks

    weights

    The vector of weights

    intercept

    The intercept (bias) weight

  6. trait WithParameters extends AnyRef

    Permalink

    Adds a ParameterMap which can be used to store configuration values

Value Members

  1. object FlinkMLTools

    Permalink

    FlinkMLTools contains a set of convenience functions for Flink's machine learning library:

    FlinkMLTools contains a set of convenience functions for Flink's machine learning library:

    • persist: Takes up to 5 DataSets and file paths. Each DataSet is written to the specified path and subsequently re-read from disk. This method can be used to effectively split the execution graph at the given DataSet. Writing it to disk triggers its materialization and specifying it as a source will prevent the re-execution of it.
    • block: Takes a DataSet of elements T and groups them in n blocks.
  2. object ParameterMap extends Serializable

    Permalink

Ungrouped