de.jstacs.algorithms.graphs.tensor
Class SymmetricTensor

java.lang.Object
  de.jstacs.algorithms.graphs.tensor.Tensor
      de.jstacs.algorithms.graphs.tensor.SymmetricTensor

public class SymmetricTensor
extends Tensor

This class can be used for Tensors with a special symmetry property. This property states that for a given target node and for all permutations of parent nodes the edge weight is constant. This is very helpful to save storage and time in some algorithms. Since it is very useful it also enables you to use this class if your tensor does not fulfill this property. For that case the tensor saves only the maximum edge weight and the permutation of the parent nodes.

Author:

Jens Keilwagen

See Also:

Tensor, AsymmetricTensor

Field Summary

Fields inherited from class de.jstacs.algorithms.graphs.tensor.Tensor

L, order, powers

Constructor Summary

SymmetricTensor(AsymmetricTensor asym_tensor)
This constructor creates and checks a filled asymmetric tensor from an AsymmetricTensor instance.

SymmetricTensor(double[][][] asym_tensor, int N, byte k)
This constructor creates and checks a filled asymmetric tensor with given dimension.

SymmetricTensor(int n, byte k)
This constructor creates an empty symmetric tensor with given dimension.

SymmetricTensor(SymmetricTensor[] parts, double[] weights)
The constructor can be used creating a new SymmetricTensor as weighted sum of SymmetricTensors.

Method Summary

double	getBest(int child, int[] par, byte k) Returns the maximal weight which can be reached for an edge from k nodes from the (encoded) set par to the node child.
int[]	getEdgeFromIndex(int idx, int k) This method decodes an index in an edge.
int[]	getMaximalEdgeFor(byte k, int child, int... parents) Returns the edge permute(parents[0],...
double	getRootValue(int child) Returns the value for child as root.
int	getTrueIndexForLastGetBest() Returns the edge from getBest(int, int[], byte) in an encoded index.
double	getValue(byte k, int child, int... parents) Returns the value for the edge parents[0],...
void	resetValue(byte k, int child, int... parents) Sets the value for the edge parents -> child to Double.NEGATIVE_INFTINITY.
void	setRootValue(int child, double val) Sets the value val for the root node child.
void	setValue(byte k, double val, int child, int... parents) Sets the value if it is bigger than the current value and keeps the parents information.

Methods inherited from class de.jstacs.algorithms.graphs.tensor.Tensor

getAsymIndex, getNumberOfNodes, getOrder, readTensorFromFile, toDouble3DArray, writeTensorToFile

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SymmetricTensor

public SymmetricTensor(int n,
                       byte k)

This constructor creates an empty symmetric tensor with given dimension.

Parameters:

n - the number of nodes

k - the order

See Also:

Tensor.Tensor(int, byte)

SymmetricTensor

public SymmetricTensor(SymmetricTensor[] parts,
                       double[] weights)
                throws IllegalArgumentException

The constructor can be used creating a new SymmetricTensor as weighted sum of SymmetricTensors.

Parameters:

parts - the SymmetricTensors

weights - the weight for each SymmetricTensor

Throws:

IllegalArgumentException - if something went wrong

See Also:

SymmetricTensor(int, byte)

SymmetricTensor

public SymmetricTensor(AsymmetricTensor asym_tensor)

This constructor creates and checks a filled asymmetric tensor from an AsymmetricTensor instance.

Parameters:

asym_tensor - the asymmetric instance

See Also:

AsymmetricTensor, SymmetricTensor(double[][][], int, byte)

SymmetricTensor

public SymmetricTensor(double[][][] asym_tensor,
                       int N,
                       byte k)

This constructor creates and checks a filled asymmetric tensor with given dimension.

Parameters:

asym_tensor - the tensor weights

N - the number of nodes

k - the order

See Also:

SymmetricTensor(int, byte)

Method Detail

getBest

public double getBest(int child,
                      int[] par,
                      byte k)

Returns the maximal weight which can be reached for an edge from k nodes from the (encoded) set par to the node child.

Parameters:

child - the target node

par - an encoded set, elements are sorted and for all elements e >= child the real value is (e+1), e.g. child = 3 and possible parents = {5,2,0,6,1} then par = {0,1,2,4,5}

k - the number of parents used from par

Returns:

the best (=maximal) weight

getEdgeFromIndex

public int[] getEdgeFromIndex(int idx,
                              int k)

This method decodes an index in an edge.

Parameters:

idx - the encoded index

k - the order

Returns:

the edge

getMaximalEdgeFor

public int[] getMaximalEdgeFor(byte k,
                               int child,
                               int... parents)

Description copied from class: Tensor

Returns the edge permute(parents[0],...,parents[k-1]) -> child that maximizes the score.

Specified by:

getMaximalEdgeFor in class Tensor

Parameters:

k - the number of parents to be used

child - the child node

parents - the parent nodes (only the first k will be used)

Returns:

the edge permute(parents[0],...,parents[k-1]) -> child that maximizes the score

getRootValue

public double getRootValue(int child)

Description copied from class: Tensor

Returns the value for child as root.

Specified by:

getRootValue in class Tensor

Parameters:

child - the name of the node

Returns:

the value for the node

getTrueIndexForLastGetBest

public int getTrueIndexForLastGetBest()

Returns the edge from getBest(int, int[], byte) in an encoded index.

Returns:

the edge from getBest(int, int[], byte) in an encoded index

See Also:

getBest(int, int[], byte)

getValue

public double getValue(byte k,
                       int child,
                       int... parents)

Description copied from class: Tensor

Returns the value for the edge parents[0],...,parents[k-1] -> child.

Specified by:

getValue in class Tensor

Parameters:

k - the number of parents to be used

child - the child node

parents - the parent nodes (only the first k will be used)

Returns:

the value for the edge parents[0],...,parents[k-1] -> child

resetValue

public void resetValue(byte k,
                       int child,
                       int... parents)

Sets the value for the edge parents -> child to Double.NEGATIVE_INFTINITY.

Specified by:

resetValue in class Tensor

Parameters:

k - the order

child - the index of the child node

parents - the indices of the parent nodes

setRootValue

public void setRootValue(int child,
                         double val)

Description copied from class: Tensor

Sets the value val for the root node child.

Specified by:

setRootValue in class Tensor

Parameters:

child - the name of the node

val - the value for the node

setValue

public void setValue(byte k,
                     double val,
                     int child,
                     int... parents)

Sets the value if it is bigger than the current value and keeps the parents information.

Specified by:

setValue in class Tensor

Parameters:

k - the order

val - the value to be set

child - the index of the child node

parents - the indices of the parent nodes