PersonalizedPageRankAnchorSearcher

class PersonalizedPageRankAnchorSearcher(batch_size=1, use_tqdm=False, page_rank_kwargs=None)[source]

Bases: pykeen.nn.node_piece.anchor_search.AnchorSearcher

Select closest anchors as the nodes with the largest personalized page rank.

See also

http://web.stanford.edu/class/cs224w/slides/04-pagerank.pdf

Initialize the searcher.

Parameters

  • batch_size (int) – the batch size to use.

  • use_tqdm (bool) – whether to use tqdm

  • page_rank_kwargs (Optional[Mapping[str, Any]]) – keyword-based parameters used for page_rank(). Must not include edge_index, or x0.

Methods Summary

__call__(edge_index, anchors, k)

Find the \(k\) closest anchor nodes for each entity.

extra_repr()

Extra components for __repr__.

precalculate_anchor_ppr(edge_index, anchors)

Sort anchors nodes by PPR values from each node.

Methods Documentation

__call__(edge_index, anchors, k)[source]

Find the \(k\) closest anchor nodes for each entity.

Parameters

  • edge_index (ndarray) – shape: (2, m) the edge index

  • anchors (ndarray) – shape: (a,) the selected anchor entity Ids

  • k (int) – the number of closest anchors to return

Return type

ndarray

Returns

shape: (n, k), -1 <= res < a the Ids of the closest anchors

extra_repr()[source]

Extra components for __repr__.

Return type

Iterable[str]

precalculate_anchor_ppr(edge_index, anchors)[source]

Sort anchors nodes by PPR values from each node.

Parameters

  • edge_index (ndarray) – shape: (2, m) the edge index.

  • anchors (ndarray) – shape: (num_anchors,) the anchor IDs.

Return type

ndarray

Returns

shape: (num_entities, num_anchors) the PPR values for each anchor