Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313179953> ?p ?o ?g. }
Showing items 1 to 50 of
50
with 100 items per page.
- W4313179953 abstract "Recently, Graph Neural Networks (GNNs) have been focused because they are suitable for considering complex interactions. Such a real-world problem containing complex interactions can be represented by graph-structured data. However, this graph-structured data may contain task-irrelevant edges or missing edges due to human error. These noisy or incomplete edges can degrade the performance of trained GNNs. To tackle this problem, Graph Structure Learning (GSL), which focuses on finding a better graph representation of a given original graph based on the main learning objectives, has been studied. In this research, we propose a framework called Confusion-GSL that includes conventional graph neural network for node classification and edge selection for finding better graph representation. We focus on the fact that the probability matrix can be used to reconstruct graph representation like learned node embedding in Graph Auto Encoder (GAE). After training node classification, we calculate Gram matrix of the probability matrix and use this Gram matrix as an approximation of adjacency matrix. By applying Gumbel-Softmax to difference matrix between approximated adjacency matrix and original adjacency matrix, we select the connections likely to exist or not exist. After that we re-train GNN with this modified graph representation. We evaluate our proposed framework on two citation network datasets commonly used: Cora, Citeseer and Pubmed." @default.
- W4313179953 created "2023-01-06" @default.
- W4313179953 creator A5033518314 @default.
- W4313179953 creator A5082598779 @default.
- W4313179953 date "2022-07-05" @default.
- W4313179953 modified "2023-09-27" @default.
- W4313179953 title "Graph Structure Learning based on Mistakenly Predicted Edges from Reconstructed Graph Representation" @default.
- W4313179953 cites W3035467734 @default.
- W4313179953 cites W3116239416 @default.
- W4313179953 doi "https://doi.org/10.1109/itc-cscc55581.2022.9894945" @default.
- W4313179953 hasPublicationYear "2022" @default.
- W4313179953 type Work @default.
- W4313179953 citedByCount "1" @default.
- W4313179953 countsByYear W43131799532022 @default.
- W4313179953 crossrefType "proceedings-article" @default.
- W4313179953 hasAuthorship W4313179953A5033518314 @default.
- W4313179953 hasAuthorship W4313179953A5082598779 @default.
- W4313179953 hasConcept C132525143 @default.
- W4313179953 hasConcept C154945302 @default.
- W4313179953 hasConcept C180356752 @default.
- W4313179953 hasConcept C203776342 @default.
- W4313179953 hasConcept C22149727 @default.
- W4313179953 hasConcept C41008148 @default.
- W4313179953 hasConcept C78913703 @default.
- W4313179953 hasConcept C80444323 @default.
- W4313179953 hasConceptScore W4313179953C132525143 @default.
- W4313179953 hasConceptScore W4313179953C154945302 @default.
- W4313179953 hasConceptScore W4313179953C180356752 @default.
- W4313179953 hasConceptScore W4313179953C203776342 @default.
- W4313179953 hasConceptScore W4313179953C22149727 @default.
- W4313179953 hasConceptScore W4313179953C41008148 @default.
- W4313179953 hasConceptScore W4313179953C78913703 @default.
- W4313179953 hasConceptScore W4313179953C80444323 @default.
- W4313179953 hasFunder F4320324891 @default.
- W4313179953 hasLocation W43131799531 @default.
- W4313179953 hasOpenAccess W4313179953 @default.
- W4313179953 hasPrimaryLocation W43131799531 @default.
- W4313179953 hasRelatedWork W2021356981 @default.
- W4313179953 hasRelatedWork W2058863301 @default.
- W4313179953 hasRelatedWork W2563497088 @default.
- W4313179953 hasRelatedWork W267237930 @default.
- W4313179953 hasRelatedWork W2928010092 @default.
- W4313179953 hasRelatedWork W3170778148 @default.
- W4313179953 hasRelatedWork W4296210796 @default.
- W4313179953 hasRelatedWork W4301033264 @default.
- W4313179953 hasRelatedWork W4318995835 @default.
- W4313179953 hasRelatedWork W4320342417 @default.
- W4313179953 isParatext "false" @default.
- W4313179953 isRetracted "false" @default.
- W4313179953 workType "article" @default.