org.deeplearning4j.graph.iterator

Class WeightedRandomWalkIterator<V>

java.lang.Object

org.deeplearning4j.graph.iterator.WeightedRandomWalkIterator<V>

All Implemented Interfaces:

GraphWalkIterator<V>

public class WeightedRandomWalkIterator<V>
extends java.lang.Object
implements GraphWalkIterator<V>

Given a graph, iterate through random walks on that graph of a specified length. Unlike RandomWalkIterator, the WeightedRandomWalkIterator uses the values associated with each edge to determine probabilities. Weights on each edge need not be normalized.
Because the edge values are used to determine the probabilities of selecting an edge, the WeightedRandomWalkIterator can only be used on graphs with an edge type that extends the Number class (i.e., Integer, Double, etc)
Random walks are generated starting at every node in the graph exactly once, though the order of the starting nodes is randomized.

Constructor Summary

Constructors

Constructor and Description
WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph, int walkLength)
WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph, int walkLength, long rngSeed) Construct a RandomWalkIterator for a given graph, with a specified walk length and random number generator seed. Uses NoEdgeHandling.EXCEPTION_ON_DISCONNECTED - hence exception will be thrown when generating random walks on graphs with vertices containing having no edges, or no outgoing edges (for directed graphs)
WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph, int walkLength, long rngSeed, NoEdgeHandling mode)
WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph, int walkLength, long rngSeed, NoEdgeHandling mode, int firstVertex, int lastVertex) Constructor used to generate random walks starting at a subset of the vertices in the graph.

Method Summary

Modifier and Type	Method and Description
boolean	hasNext() Whether the iterator has any more vertex sequences.
IVertexSequence<V>	next() Get the next vertex sequence.
void	reset() Reset the graph walk iterator.
int	walkLength() Length of the walks returned by next() Note that a walk of length i contains i+1 vertices

Methods inherited from class java.lang.Object

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

Constructor Detail

WeightedRandomWalkIterator

public WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph,
                                  int walkLength)

WeightedRandomWalkIterator

public WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph,
                                  int walkLength,
                                  long rngSeed)

Construct a RandomWalkIterator for a given graph, with a specified walk length and random number generator seed.
Uses NoEdgeHandling.EXCEPTION_ON_DISCONNECTED - hence exception will be thrown when generating random walks on graphs with vertices containing having no edges, or no outgoing edges (for directed graphs)

See Also:

WeightedRandomWalkIterator(IGraph, int, long, NoEdgeHandling)

WeightedRandomWalkIterator

public WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph,
                                  int walkLength,
                                  long rngSeed,
                                  NoEdgeHandling mode)

Parameters:

graph - IGraph to conduct walks on

walkLength - length of each walk. Walk of length 0 includes 1 vertex, walk of 1 includes 2 vertices etc

rngSeed - seed for randomization

mode - mode for handling random walks from vertices with either no edges, or no outgoing edges (for directed graphs)

WeightedRandomWalkIterator

public WeightedRandomWalkIterator(IGraph<V,? extends java.lang.Number> graph,
                                  int walkLength,
                                  long rngSeed,
                                  NoEdgeHandling mode,
                                  int firstVertex,
                                  int lastVertex)

Constructor used to generate random walks starting at a subset of the vertices in the graph. Order of starting vertices is randomized within this subset

Parameters:

graph - IGraph to conduct walks on

walkLength - length of each walk. Walk of length 0 includes 1 vertex, walk of 1 includes 2 vertices etc

rngSeed - seed for randomization

mode - mode for handling random walks from vertices with either no edges, or no outgoing edges (for directed graphs)

firstVertex - first vertex index (inclusive) to start random walks from

lastVertex - last vertex index (exclusive) to start random walks from

Method Detail

next

public IVertexSequence<V> next()

Description copied from interface: GraphWalkIterator

Get the next vertex sequence.

Specified by:

next in interface GraphWalkIterator<V>

hasNext

public boolean hasNext()

Description copied from interface: GraphWalkIterator

Whether the iterator has any more vertex sequences.

Specified by:

hasNext in interface GraphWalkIterator<V>

reset

public void reset()

Description copied from interface: GraphWalkIterator

Reset the graph walk iterator.

Specified by:

reset in interface GraphWalkIterator<V>

walkLength

public int walkLength()

Description copied from interface: GraphWalkIterator

Length of the walks returned by next() Note that a walk of length i contains i+1 vertices

Specified by:

walkLength in interface GraphWalkIterator<V>