Density-based Neural Temporal Point Processes for Heartbeat Dynamics

TL;DR

This paper introduces a density-based neural temporal point process model for heartbeat dynamics, demonstrating its effectiveness in capturing physiological patterns and enabling zero-shot predictions from heartbeat data.

Contribution

It adapts classical goodness-of-fit methods to neural TPPs for optimizing hyperparameters and sequence lengths, advancing heartbeat modeling techniques.

Findings

Effective modeling of heartbeat dynamics across 18 subjects.

Successful zero-shot prediction of heartbeat sequences.

Optimized hyperparameters improve model performance.

Abstract

Temporal point processes (TPPs) provide a natural mathematical framework for modeling heartbeats due to capturing underlying physiological inductive biases. In this work, we apply density-based neural TPPs to model heartbeat dynamics from 18 subjects. We adapt a goodness-of-fit framework from classical point process literature to Neural TPPs and use it to optimize hyperparameters, identify appropriate training sequence lengths to capture temporal dependencies, and demonstrate zero-shot predictive capability on heartbeat data.