SGCL replaces the two-loss multi-task setup in graph recommendation with one supervised contrastive loss, reporting better accuracy and speed on Beauty and Toys-and-Games.
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
-
SGCL: Unifying Self-Supervised and Supervised Learning for Graph Recommendation
SGCL replaces the two-loss multi-task setup in graph recommendation with one supervised contrastive loss, reporting better accuracy and speed on Beauty and Toys-and-Games.