Matches in SemOpenAlex for { <https://semopenalex.org/work/W3213730107> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W3213730107 abstract "In recent years, the Transformer architecture has proven to be very successful in sequence processing, but its application to other data structures, such as graphs, has remained limited due to the difficulty of properly defining positions. Here, we present the $textit{Spectral Attention Network}$ (SAN), which uses a learned positional encoding (LPE) that can take advantage of the full Laplacian spectrum to learn the position of each node in a given graph. This LPE is then added to the node features of the graph and passed to a fully-connected Transformer. By leveraging the full spectrum of the Laplacian, our model is theoretically powerful in distinguishing graphs, and can better detect similar sub-structures from their resonance. Further, by fully connecting the graph, the Transformer does not suffer from over-squashing, an information bottleneck of most GNNs, and enables better modeling of physical phenomenons such as heat transfer and electric interaction. When tested empirically on a set of 4 standard datasets, our model performs on par or better than state-of-the-art GNNs, and outperforms any attention-based model by a wide margin, becoming the first fully-connected architecture to perform well on graph benchmarks." @default.
- W3213730107 created "2021-11-22" @default.
- W3213730107 creator A5004542284 @default.
- W3213730107 creator A5037753193 @default.
- W3213730107 creator A5064214996 @default.
- W3213730107 creator A5074836852 @default.
- W3213730107 creator A5083336206 @default.
- W3213730107 date "2021-12-06" @default.
- W3213730107 modified "2023-09-24" @default.
- W3213730107 title "Rethinking Graph Transformers with Spectral Attention" @default.
- W3213730107 hasPublicationYear "2021" @default.
- W3213730107 type Work @default.
- W3213730107 sameAs 3213730107 @default.
- W3213730107 citedByCount "1" @default.
- W3213730107 countsByYear W32137301072021 @default.
- W3213730107 crossrefType "proceedings-article" @default.
- W3213730107 hasAuthorship W3213730107A5004542284 @default.
- W3213730107 hasAuthorship W3213730107A5037753193 @default.
- W3213730107 hasAuthorship W3213730107A5064214996 @default.
- W3213730107 hasAuthorship W3213730107A5074836852 @default.
- W3213730107 hasAuthorship W3213730107A5083336206 @default.
- W3213730107 hasConcept C11413529 @default.
- W3213730107 hasConcept C115178988 @default.
- W3213730107 hasConcept C119599485 @default.
- W3213730107 hasConcept C127413603 @default.
- W3213730107 hasConcept C132525143 @default.
- W3213730107 hasConcept C134306372 @default.
- W3213730107 hasConcept C149635348 @default.
- W3213730107 hasConcept C154945302 @default.
- W3213730107 hasConcept C165700671 @default.
- W3213730107 hasConcept C165801399 @default.
- W3213730107 hasConcept C2780513914 @default.
- W3213730107 hasConcept C33923547 @default.
- W3213730107 hasConcept C41008148 @default.
- W3213730107 hasConcept C66322947 @default.
- W3213730107 hasConcept C80444323 @default.
- W3213730107 hasConceptScore W3213730107C11413529 @default.
- W3213730107 hasConceptScore W3213730107C115178988 @default.
- W3213730107 hasConceptScore W3213730107C119599485 @default.
- W3213730107 hasConceptScore W3213730107C127413603 @default.
- W3213730107 hasConceptScore W3213730107C132525143 @default.
- W3213730107 hasConceptScore W3213730107C134306372 @default.
- W3213730107 hasConceptScore W3213730107C149635348 @default.
- W3213730107 hasConceptScore W3213730107C154945302 @default.
- W3213730107 hasConceptScore W3213730107C165700671 @default.
- W3213730107 hasConceptScore W3213730107C165801399 @default.
- W3213730107 hasConceptScore W3213730107C2780513914 @default.
- W3213730107 hasConceptScore W3213730107C33923547 @default.
- W3213730107 hasConceptScore W3213730107C41008148 @default.
- W3213730107 hasConceptScore W3213730107C66322947 @default.
- W3213730107 hasConceptScore W3213730107C80444323 @default.
- W3213730107 hasLocation W32137301071 @default.
- W3213730107 hasOpenAccess W3213730107 @default.
- W3213730107 hasPrimaryLocation W32137301071 @default.
- W3213730107 hasRelatedWork W2784304320 @default.
- W3213730107 hasRelatedWork W2899886700 @default.
- W3213730107 hasRelatedWork W2912077575 @default.
- W3213730107 hasRelatedWork W2935184916 @default.
- W3213730107 hasRelatedWork W3015103993 @default.
- W3213730107 hasRelatedWork W3081636763 @default.
- W3213730107 hasRelatedWork W3093901489 @default.
- W3213730107 hasRelatedWork W3102154670 @default.
- W3213730107 hasRelatedWork W3147316959 @default.
- W3213730107 hasRelatedWork W3163139382 @default.
- W3213730107 hasRelatedWork W3169575312 @default.
- W3213730107 hasRelatedWork W3184380725 @default.
- W3213730107 hasRelatedWork W3185051168 @default.
- W3213730107 hasRelatedWork W3189076644 @default.
- W3213730107 hasRelatedWork W3191962800 @default.
- W3213730107 hasRelatedWork W3196983225 @default.
- W3213730107 hasRelatedWork W3199120587 @default.
- W3213730107 hasRelatedWork W3205017380 @default.
- W3213730107 hasRelatedWork W3209214366 @default.
- W3213730107 hasRelatedWork W3210288927 @default.
- W3213730107 hasVolume "34" @default.
- W3213730107 isParatext "false" @default.
- W3213730107 isRetracted "false" @default.
- W3213730107 magId "3213730107" @default.
- W3213730107 workType "article" @default.