Speculative Sampling for Parametric Temporal Point Processes
This addresses a bottleneck in large-scale TPP applications by bridging the gap between expressive modeling and efficient parallel generation.
The paper tackles the problem of inefficient sequential sampling in temporal point processes by introducing a rejection sampling algorithm that enables exact parallel generation of multiple future events without model changes or retraining, achieving empirical speedups on real-world datasets.
Temporal point processes are powerful generative models for event sequences that capture complex dependencies in time-series data. They are commonly specified using autoregressive models that learn the distribution of the next event from the previous events. This makes sampling inherently sequential, limiting efficiency. In this paper, we propose a novel algorithm based on rejection sampling that enables exact sampling of multiple future values from existing TPP models, in parallel, and without requiring any architectural changes or retraining. Besides theoretical guarantees, our method demonstrates empirical speedups on real-world datasets, bridging the gap between expressive modeling and efficient parallel generation for large-scale TPP applications.