Class NeuronSquareMesh2D

java.lang.Object
org.apache.commons.math3.ml.neuralnet.twod.NeuronSquareMesh2D
All Implemented Interfaces:
Serializable, Iterable<Neuron>

public class NeuronSquareMesh2D extends Object implements Iterable<Neuron>, Serializable
Neural network with the topology of a two-dimensional surface. Each neuron defines one surface element.
This network is primarily intended to represent a Self Organizing Feature Map.
Since:
3.3
See Also:
  • Field Details

    • serialVersionUID

      private static final long serialVersionUID
      Serial version ID
      See Also:
    • network

      private final Network network
      Underlying network.
    • numberOfRows

      private final int numberOfRows
      Number of rows.
    • numberOfColumns

      private final int numberOfColumns
      Number of columns.
    • wrapRows

      private final boolean wrapRows
      Wrap.
    • wrapColumns

      private final boolean wrapColumns
      Wrap.
    • neighbourhood

      private final SquareNeighbourhood neighbourhood
      Neighbourhood type.
    • identifiers

      private final long[][] identifiers
      Mapping of the 2D coordinates (in the rectangular mesh) to the neuron identifiers (attributed by the network instance).
  • Constructor Details

    • NeuronSquareMesh2D

      NeuronSquareMesh2D(boolean wrapRowDim, boolean wrapColDim, SquareNeighbourhood neighbourhoodType, double[][][] featuresList)
      Constructor with restricted access, solely used for deserialization.
      Parameters:
      wrapRowDim - Whether to wrap the first dimension (i.e the first and last neurons will be linked together).
      wrapColDim - Whether to wrap the second dimension (i.e the first and last neurons will be linked together).
      neighbourhoodType - Neighbourhood type.
      featuresList - Arrays that will initialize the features sets of the network's neurons.
      Throws:
      NumberIsTooSmallException - if numRows < 2 or numCols < 2.
    • NeuronSquareMesh2D

      public NeuronSquareMesh2D(int numRows, boolean wrapRowDim, int numCols, boolean wrapColDim, SquareNeighbourhood neighbourhoodType, FeatureInitializer[] featureInit)
      Creates a two-dimensional network composed of square cells: Each neuron not located on the border of the mesh has four neurons linked to it.
      The links are bi-directional.
      The topology of the network can also be a cylinder (if one of the dimensions is wrapped) or a torus (if both dimensions are wrapped).
      Parameters:
      numRows - Number of neurons in the first dimension.
      wrapRowDim - Whether to wrap the first dimension (i.e the first and last neurons will be linked together).
      numCols - Number of neurons in the second dimension.
      wrapColDim - Whether to wrap the second dimension (i.e the first and last neurons will be linked together).
      neighbourhoodType - Neighbourhood type.
      featureInit - Array of functions that will initialize the corresponding element of the features set of each newly created neuron. In particular, the size of this array defines the size of feature set.
      Throws:
      NumberIsTooSmallException - if numRows < 2 or numCols < 2.
    • NeuronSquareMesh2D

      private NeuronSquareMesh2D(boolean wrapRowDim, boolean wrapColDim, SquareNeighbourhood neighbourhoodType, Network net, long[][] idGrid)
      Constructor with restricted access, solely used for making a deep copy.
      Parameters:
      wrapRowDim - Whether to wrap the first dimension (i.e the first and last neurons will be linked together).
      wrapColDim - Whether to wrap the second dimension (i.e the first and last neurons will be linked together).
      neighbourhoodType - Neighbourhood type.
      net - Underlying network.
      idGrid - Neuron identifiers.
  • Method Details

    • copy

      public NeuronSquareMesh2D copy()
      Performs a deep copy of this instance. Upon return, the copied and original instances will be independent: Updating one will not affect the other.
      Returns:
      a new instance with the same state as this instance.
      Since:
      3.6
    • iterator

      public Iterator<Neuron> iterator()
      Specified by:
      iterator in interface Iterable<Neuron>
      Since:
      3.6
    • getNetwork

      public Network getNetwork()
      Retrieves the underlying network. A reference is returned (enabling, for example, the network to be trained). This also implies that calling methods that modify the Network topology may cause this class to become inconsistent.
      Returns:
      the network.
    • getNumberOfRows

      public int getNumberOfRows()
      Gets the number of neurons in each row of this map.
      Returns:
      the number of rows.
    • getNumberOfColumns

      public int getNumberOfColumns()
      Gets the number of neurons in each column of this map.
      Returns:
      the number of column.
    • getNeuron

      public Neuron getNeuron(int i, int j)
      Retrieves the neuron at location (i, j) in the map. The neuron at position (0, 0) is located at the upper-left corner of the map.
      Parameters:
      i - Row index.
      j - Column index.
      Returns:
      the neuron at (i, j).
      Throws:
      OutOfRangeException - if i or j is out of range.
      See Also:
    • getNeuron

      public Neuron getNeuron(int row, int col, NeuronSquareMesh2D.HorizontalDirection alongRowDir, NeuronSquareMesh2D.VerticalDirection alongColDir)
      Retrieves the neuron at (location[0], location[1]) in the map. The neuron at position (0, 0) is located at the upper-left corner of the map.
      Parameters:
      row - Row index.
      col - Column index.
      alongRowDir - Direction along the given row (i.e. an offset will be added to the given column index.
      alongColDir - Direction along the given col (i.e. an offset will be added to the given row index.
      Returns:
      the neuron at the requested location, or null if the location is not on the map.
      See Also:
    • getLocation

      private int[] getLocation(int row, int col, NeuronSquareMesh2D.HorizontalDirection alongRowDir, NeuronSquareMesh2D.VerticalDirection alongColDir)
      Computes the location of a neighbouring neuron. It will return null if the resulting location is not part of the map. Position (0, 0) is at the upper-left corner of the map.
      Parameters:
      row - Row index.
      col - Column index.
      alongRowDir - Direction along the given row (i.e. an offset will be added to the given column index.
      alongColDir - Direction along the given col (i.e. an offset will be added to the given row index.
      Returns:
      an array of length 2 containing the indices of the requested location, or null if that location is not part of the map.
      See Also:
    • createLinks

      private void createLinks()
      Creates the neighbour relationships between neurons.
    • readObject

      private void readObject(ObjectInputStream in)
      Prevents proxy bypass.
      Parameters:
      in - Input stream.
    • writeReplace

      private Object writeReplace()
      Custom serialization.
      Returns:
      the proxy instance that will be actually serialized.