Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387378554> ?p ?o ?g. }
Showing items 1 to 60 of
60
with 100 items per page.
- W4387378554 endingPage "126850" @default.
- W4387378554 startingPage "126850" @default.
- W4387378554 abstract "Contrastive learning has recently achieved remarkable success in many domains including graphs. However contrastive loss, especially for graphs, requires a large number of negative samples which is unscalable and computationally prohibitive with a quadratic time complexity. Sub-sampling is not optimal. Incorrect negative sampling leads to sampling bias. In this work, we propose a meta-node based approximation technique that is (a) simple, (b) canproxy all negative combinations (c) in quadratic cluster size time complexity, (d) at graph level, not node level, and (e) exploit graph sparsity. By replacing node-pairs with additive cluster-pairs, we compute the negatives in cluster-time at graph level. The resulting Proxy approximated meta-node Contrastive (PamC) loss, based on simple optimized GPU operations, captures the full set of negatives, yet is efficient with a linear time complexity. By avoiding sampling, we effectively eliminate sample bias. We meet the criterion for larger number of samples, thus achieving block-contrastiveness, which is proven to outperform pair-wise losses. We use learnt soft cluster assignments for the meta-node construction, and avoid possible heterophily and noise added during edge creation. Theoretically, we show that real world graphs easily satisfy conditions necessary for our approximation. Empirically, we show promising accuracy gains over state-of-the-art graph clustering on 6 benchmarks. Importantly, we gain substantially in efficiency; over 2x reduction in training time and over 5x in GPU memory reduction. Additionally, our embeddings, combined with a single learnt linear transformation, is sufficient for node classification; we achieve state-of-the-art on Citeseer classification benchmark. code:https://github.com/gayanku/PAMC" @default.
- W4387378554 created "2023-10-06" @default.
- W4387378554 creator A5040887350 @default.
- W4387378554 creator A5078468070 @default.
- W4387378554 creator A5083545460 @default.
- W4387378554 date "2023-10-01" @default.
- W4387378554 modified "2023-10-11" @default.
- W4387378554 title "Efficient block contrastive learning via parameter-free meta-node approximation" @default.
- W4387378554 cites W2100495367 @default.
- W4387378554 doi "https://doi.org/10.1016/j.neucom.2023.126850" @default.
- W4387378554 hasPublicationYear "2023" @default.
- W4387378554 type Work @default.
- W4387378554 citedByCount "0" @default.
- W4387378554 crossrefType "journal-article" @default.
- W4387378554 hasAuthorship W4387378554A5040887350 @default.
- W4387378554 hasAuthorship W4387378554A5078468070 @default.
- W4387378554 hasAuthorship W4387378554A5083545460 @default.
- W4387378554 hasBestOaLocation W43873785541 @default.
- W4387378554 hasConcept C11413529 @default.
- W4387378554 hasConcept C127413603 @default.
- W4387378554 hasConcept C129844170 @default.
- W4387378554 hasConcept C132525143 @default.
- W4387378554 hasConcept C179799912 @default.
- W4387378554 hasConcept C2524010 @default.
- W4387378554 hasConcept C311688 @default.
- W4387378554 hasConcept C33923547 @default.
- W4387378554 hasConcept C41008148 @default.
- W4387378554 hasConcept C62611344 @default.
- W4387378554 hasConcept C66938386 @default.
- W4387378554 hasConcept C80444323 @default.
- W4387378554 hasConceptScore W4387378554C11413529 @default.
- W4387378554 hasConceptScore W4387378554C127413603 @default.
- W4387378554 hasConceptScore W4387378554C129844170 @default.
- W4387378554 hasConceptScore W4387378554C132525143 @default.
- W4387378554 hasConceptScore W4387378554C179799912 @default.
- W4387378554 hasConceptScore W4387378554C2524010 @default.
- W4387378554 hasConceptScore W4387378554C311688 @default.
- W4387378554 hasConceptScore W4387378554C33923547 @default.
- W4387378554 hasConceptScore W4387378554C41008148 @default.
- W4387378554 hasConceptScore W4387378554C62611344 @default.
- W4387378554 hasConceptScore W4387378554C66938386 @default.
- W4387378554 hasConceptScore W4387378554C80444323 @default.
- W4387378554 hasLocation W43873785541 @default.
- W4387378554 hasOpenAccess W4387378554 @default.
- W4387378554 hasPrimaryLocation W43873785541 @default.
- W4387378554 hasRelatedWork W1589381382 @default.
- W4387378554 hasRelatedWork W2041119006 @default.
- W4387378554 hasRelatedWork W2320611895 @default.
- W4387378554 hasRelatedWork W2371283136 @default.
- W4387378554 hasRelatedWork W2374502283 @default.
- W4387378554 hasRelatedWork W2519362558 @default.
- W4387378554 hasRelatedWork W2922314686 @default.
- W4387378554 hasRelatedWork W3159754125 @default.
- W4387378554 hasRelatedWork W4292026839 @default.
- W4387378554 hasRelatedWork W961142965 @default.
- W4387378554 isParatext "false" @default.
- W4387378554 isRetracted "false" @default.
- W4387378554 workType "article" @default.