The Limits of Graph Samplers for Training Inductive Recommender Systems
Summary: Empirical evaluation of six graph-sampling methods on three inductive GNN recommenders and real-world datasets shows sampling can preserve accuracy using ~50% of training data while cutting training time up to 86%. Performance collapses below 50% and temporal-aware sampling is crucial, motivating new graph samplers and inductive designs. (summarized by gpt-5-mini on Feb 09 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Theis E. Jendal
- 2. Matteo Lissandrini
- 3. Peter Dolog
- 4. Katja Hose
Incoming Citations (Sorted by Pagerank)
Showing 0 of 0 citing papers.
| Rank | Citing Paper | Year | Venue | Pagerank |
|---|
Previous
Page 1 / 1
Next
Outgoing Citations (Sorted by Pagerank)
Showing 0 of 0 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|
Previous
Page 1 / 1
Next