Estimation–Prediction Tradeoff in Causal Probabilistic Temporal Graphs

arXiv·low signal

Aniq Ur Rahman argues that evaluating temporal link prediction purely by predictive accuracy on unseen edges is misleading for probabilistic temporal graphs, and formalizes an estimation–prediction tradeoff instead. The paper is relevant to teams building temporal knowledge graphs or dynamic recommendation systems who need a more principled evaluation metric. It remains largely theoretical with no released implementation noted.