EdgeRemind enables fast scalable prediction on large temporal graphs
EdgeReMIND: A Scalable, Top-Ranked Memorization Baseline for Temporal Multi-Relational Link Prediction
Machine Learning
Summary
Predicting how things connect over time on large, complex networks is hard because existing methods use too much memory or time. The authors introduce EdgeReMIND, a simple but effective model that remembers connections with a focus on relationship types. It works well on big datasets where other methods fail, achieving top accuracy while running efficiently. This shows that a straightforward memory-based approach can outperform more complicated models in real-world sized problems.
What this means in practice
- •For data engineering teams: Predict future relationships and interactions in massive network data efficiently at production scale.
- •For recommendation system developers: Build scalable recommendation engines that handle evolving user-item interactions using relation-aware memory models.
Authors
Bryant Pollard
Abstract
Temporal link prediction on the Temporal Graph Benchmark 2.0 (TGB 2.0) faces a scalability ceiling: on the benchmark's three largest datasets, every existing embedding method runs out of memory or exceeds the time budget. These large-scale graphs are the ones nearest real deployment scale, so failing on them is a real production limitation. EdgeReMIND sets the highest reported test mean reciprocal rank (MRR) on six of eight TGB 2.0 datasets and is the only relation-aware method that runs on all of them. This linear memorization model, with learned per-relation weights over data-calibrated features, is therefore not merely a fallback where embeddings fail but a practical state-of-the-art baseline across the benchmark.