Module org.tensorflow.ndarray
Package org.tensorflow.ndarray.index

Interface Index

  • public interface Index
    An index used for slicing a view out of an N-dimensional array.

    A slice, i.e. a reduced view, of an N-dimensional array is obtain by calling NdArray.slice(Index...), given a list of indices that select which elements on a given dimension should be included/excluded from that view.

    • Method Summary

      All Methods Instance Methods Abstract Methods Default Methods 
      Modifier and Type Method Description
      default Dimension apply​(Dimension dim)
      Applies this index to the given dimension.
      default long begin()
      Get the start of the index, for strided slice style indexing.
      default boolean beginMask()
      Get whether the Index should start at the beginning of the dimension, for strided slice style indexing.
      default long end()
      Get the end of the index, strided slice style indexing.
      default boolean endMask()
      Get whether the Index should end at the beginning of the dimension, for strided slice style indexing.
      default boolean isEllipsis()
      Returns true if this index is an ellipsis, expanding to take as many dimensions as possible (and applying all() to them)
      default boolean isNewAxis()
      Returns true if this index is a new axis, adding a dimension of size 1
      default boolean isPoint()
      Returns true if this index is a single point, reducing the number of dimensions by one
      default boolean isStridedSlicingCompliant()
      Get whether the Index supports strided slice style indexing (using start, end, stride, and flags, i.e.
      long mapCoordinate​(long coordinate, Dimension dim)
      Transforms an element coordinate to a new coordinate by applying this index to the given dimension.
      long numElements​(Dimension dim)
      Returns the number of elements that can be retrieved using this index on the given dimension.
      default long stride()
      Get the stride of the index, for strided slice style indexing.

    • Method Detail

      • numElements

        long numElements​(Dimension dim)
        Returns the number of elements that can be retrieved using this index on the given dimension.

        An index that maps one-by-one all elements of the dimensions will return a value equal to dim.numElements(), while an index that only maps a subset of these will return a smaller value.

        Parameters:
        dim - the indexed dimension
        Returns:
        number of elements accessible

      • mapCoordinate

        long mapCoordinate​(long coordinate,
                   Dimension dim)
        Transforms an element coordinate to a new coordinate by applying this index to the given dimension.

        For example, if the coordinate is 0 and this index flips the n elements on this dimension, then the returned value will be n-1.

        Parameters:
        coordinate - coordinate to transform
        dim - dimension the indexed dimension
        Returns:
        transformed coordinate

      • apply

        default Dimension apply​(Dimension dim)
        Applies this index to the given dimension.

        When accessing the elements from the returned dimension, this index will automatically apply and may transform the original position.

        Parameters:
        dim - dimension to apply this index to
        Returns:
        an indexed dimension

      • isPoint

        default boolean isPoint()
        Returns true if this index is a single point, reducing the number of dimensions by one

      • isNewAxis

        default boolean isNewAxis()
        Returns true if this index is a new axis, adding a dimension of size 1

      • isEllipsis

        default boolean isEllipsis()
        Returns true if this index is an ellipsis, expanding to take as many dimensions as possible (and applying all() to them)

      • isStridedSlicingCompliant

        default boolean isStridedSlicingCompliant()
        Get whether the Index supports strided slice style indexing (using start, end, stride, and flags, i.e. TensorFlow's).

      • begin

        default long begin()
        Get the start of the index, for strided slice style indexing.

      • end

        default long end()
        Get the end of the index, strided slice style indexing.

      • stride

        default long stride()
        Get the stride of the index, for strided slice style indexing.

      • beginMask

        default boolean beginMask()
        Get whether the Index should start at the beginning of the dimension, for strided slice style indexing.

      • endMask

        default boolean endMask()
        Get whether the Index should end at the beginning of the dimension, for strided slice style indexing.