WIP: Personalized PageRank #669
Conversation
|
|
||
|
|
||
| List<Node> sourceNodes = configuration.get("sourceNodes", new ArrayList<>()); | ||
| Stream<Long> sourceNodeIds = sourceNodes.stream().map(Node::getId); |
There was a problem hiding this comment.
We could use mapToLong to avoid boxing to Long instances
There was a problem hiding this comment.
Have updated this now
| concurrency, | ||
| idMapping.nodeCount(), | ||
| dampingFactor, | ||
| sourceNodeIds.map(graph::toMappedNodeId).collect(Collectors.toList()), |
There was a problem hiding this comment.
Since this is Huge, we should use toHugeMappedNodeId to get the internal long id.
There was a problem hiding this comment.
Also, a Node may not have been mapped if there was any Label/Type filtering, so maybe add a filter(mappedId -> mappedId != -1L) to remove the nodes that we don't know about.
| concurrency, | ||
| idMapping.nodeCount(), | ||
| dampingFactor, | ||
| sourceNodeIds.map(graph::toMappedNodeId).collect(Collectors.toList()), |
There was a problem hiding this comment.
If we use the LongStream here, toArray would probably better.
| .toArray(); | ||
|
|
||
| for (long sourceNodeId : partitionSourceNodeIds) { | ||
| partitionRank[Math.toIntExact(sourceNodeId - this.startNode)] = alpha; |
There was a problem hiding this comment.
@knutwalker is this the correct way to find the place in the array to update?
There was a problem hiding this comment.
@mneedham yes, that's fine; we usually have (int) (nodeId - startNode) for local nodes or (int) (nodeId - starts[partitionIndex]) for nodes in a different partition.
|
LGTM |
* test showing just normal PageRank * PPR test WIP * wip * more wip * add calculation to huge as well * this factory doesn't seem to listen to my undirected suggestion * Fixing boxing * fix imports * fixing a bug in where we set alpha. Offset should be based on our startNode * filter unmapped nodes * updating docs to mention PPR * oops we don't need this
2 participants
No description provided.