Innovation
Methods, ideas, or system contributions that make the work stand out.
Temporal Link Prediction
Source Node Activity Modeling
Negative Sampling
Point Process
Transferability
Your Temporal Link Predictor Is Blind to Who Is Active: A Missing Factor That Transfers Across Models
🤖 AI Summary
This study addresses a critical limitation in existing temporal link prediction models, where negative sampling mechanisms overlook source node activity, resulting in the loss of essential temporal signals. To mitigate this, we propose the SNAM module, which leverages self-exciting point processes and decayed interaction counts to model node activity with fewer than 20 parameters. Functioning as a plug-and-play component, SNAM enables cross-model performance transfer without retraining the backbone network. Comprehensive evaluations across 13 datasets and three evaluation protocols demonstrate that our approach achieves an overall first-place ranking, improving the Average Precision (AP) metric by up to 25 points. Furthermore, training on million-scale event streams is accelerated by 9 to 100 times compared to baselines, highlighting both computational efficiency and strong generalizability.
📝 Abstract
An interaction has two parts: someone decides to act, and then chooses whom to act on. Temporal link prediction has concentrated on the second, and we show that it is blind to the first by construction: a standard negative keeps the real source and swaps the destination, and we prove that this cancels the source's activity exactly from the optimal score, so no model trained and evaluated this way is ever rewarded for learning it. Under the harder historical and inductive negatives, whose sources differ, the same factor becomes the dominant signal. We model it with Source Node Activity Modeling (SNAM), a self-exciting event intensity fitted by an exact point-process likelihood to decayed interaction counts the history states already contain; it has fewer than 20 parameters. On their own, never looking at the destination, these parameters beat DyGFormer and TPNet on four of five datasets under historical negatives. Added to the frozen scores of TPNet, TGN, DyGFormer and DSRD, four models of different design, without retraining anything, they raise AP on almost every backbone-dataset pair in both settings, by up to 25 points. Our full model ranks first overall against eleven baselines on 13 datasets and three protocols, and on million-event streams trains an epoch 9-100x faster than TPNet and DyGFormer. We conclude that source activity is a blind spot of temporal link prediction, and a cheap, transferable one to close. Code is available at https://github.com/Erutaner/Your-Temporal-Link-Predictor-Is-Blind-to-Who-Is-Active.
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
