Package org.apache.commons.math3.ml.neuralnet

Class Network

java.lang.Object
org.apache.commons.math3.ml.neuralnet.Network
All Implemented Interfaces:
Serializable, Iterable<Neuron>
public class Network extends Object implements Iterable<Neuron>, Serializable
Neural network, composed of Neuron instances and the links between them. Although updating a neuron's state is thread-safe, modifying the network's topology (adding or removing links) is not.
Since:
3.3
See Also:

  • Field Details

  • Constructor Details

    • Network

      Network(long nextId, int featureSize, Neuron[] neuronList, long[][] neighbourIdList)
      Constructor with restricted access, solely used for deserialization.
      Parameters:
      nextId - Next available identifier.
      featureSize - Number of features.
      neuronList - Neurons.
      neighbourIdList - Links associated to each of the neurons in neuronList.
      Throws:
      MathIllegalStateException - if an inconsistency is detected (which probably means that the serialized form has been corrupted).

    • Network

      public Network(long initialIdentifier, int featureSize)
      Parameters:
      initialIdentifier - Identifier for the first neuron that will be added to this network.
      featureSize - Size of the neuron's features.

  • Method Details

    • copy

      public Network 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>

    • getNeurons

      public Collection<Neuron> getNeurons(Comparator<Neuron> comparator)
      Creates a list of the neurons, sorted in a custom order.
      Parameters:
      comparator - Comparator used for sorting the neurons.
      Returns:
      a list of neurons, sorted in the order prescribed by the given comparator.
      See Also:

    • createNeuron

      public long createNeuron(double[] features)
      Creates a neuron and assigns it a unique identifier.
      Parameters:
      features - Initial values for the neuron's features.
      Returns:
      the neuron's identifier.
      Throws:
      DimensionMismatchException - if the length of features is different from the expected size (as set by the constructor).

    • deleteNeuron

      public void deleteNeuron(Neuron neuron)
      Deletes a neuron. Links from all neighbours to the removed neuron will also be deleted.
      Parameters:
      neuron - Neuron to be removed from this network.
      Throws:
      NoSuchElementException - if n does not belong to this network.

    • getFeaturesSize

      public int getFeaturesSize()
      Gets the size of the neurons' features set.
      Returns:
      the size of the features set.

    • addLink

      public void addLink(Neuron a, Neuron b)
      Adds a link from neuron a to neuron b. Note: the link is not bi-directional; if a bi-directional link is required, an additional call must be made with a and b exchanged in the argument list.
      Parameters:
      a - Neuron.
      b - Neuron.
      Throws:
      NoSuchElementException - if the neurons do not exist in the network.

    • addLinkToLinkSet

      private void addLinkToLinkSet(Set<Long> linkSet, long id)
      Adds a link to neuron id in given linkSet. Note: no check verifies that the identifier indeed belongs to this network.
      Parameters:
      linkSet - Neuron identifier.
      id - Neuron identifier.

    • deleteLink

      public void deleteLink(Neuron a, Neuron b)
      Deletes the link between neurons a and b.
      Parameters:
      a - Neuron.
      b - Neuron.
      Throws:
      NoSuchElementException - if the neurons do not exist in the network.

    • deleteLinkFromLinkSet

      private void deleteLinkFromLinkSet(Set<Long> linkSet, long id)
      Deletes a link to neuron id in given linkSet. Note: no check verifies that the identifier indeed belongs to this network.
      Parameters:
      linkSet - Neuron identifier.
      id - Neuron identifier.

    • getNeuron

      public Neuron getNeuron(long id)
      Retrieves the neuron with the given (unique) id.
      Parameters:
      id - Identifier.
      Returns:
      the neuron associated with the given id.
      Throws:
      NoSuchElementException - if the neuron does not exist in the network.

    • getNeighbours

      public Collection<Neuron> getNeighbours(Iterable<Neuron> neurons)
      Retrieves the neurons in the neighbourhood of any neuron in the neurons list.
      Parameters:
      neurons - Neurons for which to retrieve the neighbours.
      Returns:
      the list of neighbours.
      See Also:

    • getNeighbours

      public Collection<Neuron> getNeighbours(Iterable<Neuron> neurons, Iterable<Neuron> exclude)
      Retrieves the neurons in the neighbourhood of any neuron in the neurons list. The exclude list allows to retrieve the "concentric" neighbourhoods by removing the neurons that belong to the inner "circles".
      Parameters:
      neurons - Neurons for which to retrieve the neighbours.
      exclude - Neurons to exclude from the returned list. Can be null.
      Returns:
      the list of neighbours.

    • getNeighbours

      public Collection<Neuron> getNeighbours(Neuron neuron)
      Retrieves the neighbours of the given neuron.
      Parameters:
      neuron - Neuron for which to retrieve the neighbours.
      Returns:
      the list of neighbours.
      See Also:

    • getNeighbours

      public Collection<Neuron> getNeighbours(Neuron neuron, Iterable<Neuron> exclude)
      Retrieves the neighbours of the given neuron.
      Parameters:
      neuron - Neuron for which to retrieve the neighbours.
      exclude - Neurons to exclude from the returned list. Can be null.
      Returns:
      the list of neighbours.

    • createNextId

      private Long createNextId()
      Creates a neuron identifier.
      Returns:
      a value that will serve as a unique identifier.

    • 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.