NLGCL treats each user/item embedding and its graph neighbors at the next GNN layer as positive pairs, eliminating augmentation-based contrastive views in GNN recommenders while improving accuracy and training speed.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
NLGCL: Naturally Existing Neighbor Layers Graph Contrastive Learning for Recommendation
NLGCL treats each user/item embedding and its graph neighbors at the next GNN layer as positive pairs, eliminating augmentation-based contrastive views in GNN recommenders while improving accuracy and training speed.