Package | Description |
---|---|
org.deeplearning4j.models.sequencevectors.graph.walkers.impl |
Modifier and Type | Method and Description |
---|---|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setNoEdgeHandling(@NonNull NoEdgeHandling handling)
This method defines walker behavior when it gets to node which has no next nodes available
Default value: RESTART_ON_DISCONNECTED
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setPopularityMode(@NonNull PopularityMode popularityMode)
This method defines which nodes should be taken in account when choosing next hope: maximum popularity, lowest popularity, or average popularity.
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setPopularitySpread(int topN)
This method defines, how much nodes should take place in next hop selection.
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setRestartProbability(double alpha)
This method defines a chance for walk restart
Good value would be somewhere between 0.03-0.07
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setSeed(long seed)
This method specifies random seed.
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setSpreadSpectrum(@NonNull SpreadSpectrum spectrum)
This method allows you to define, if nodes within popularity spread should have equal chances to be picked for next hop, or they should have chances proportional to their popularity.
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setWalkDirection(@NonNull WalkDirection direction)
This method defines next hop selection within walk
|
PopularityWalker.Builder<T> |
PopularityWalker.Builder.setWalkLength(int walkLength)
This method specifies output sequence (walk) length
|
Copyright © 2020. All rights reserved.