Density-based Neural Temporal Point Processes for Heartbeat Dynamics

2025-11-26

AI Summary

This work addresses the challenge of accurately characterizing temporal dependencies in heartbeat dynamics modeling. We propose a density-based neural temporal point process (Neural TPP) framework. Leveraging classical point process goodness-of-fit (GoF) testing, we automate hyperparameter optimization and adaptively determine optimal training sequence lengths—eliminating manual tuning. Our key contribution is the first extension of GoF testing to neural TPPs, enabling zero-shot, training-free heartbeat prediction and quantifying the dominant temporal dependency scale in cardiac dynamics (~5–10 seconds). Evaluated on ECG data from 18 subjects, the method significantly improves prediction accuracy and robustly uncovers intrinsic rhythmic temporal structures in physiological heartbeats.

Technology Category

• Planning, Routing, and Scheduling: Temporal Planning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Models for Web evolution

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.

Problem

• Model heartbeat dynamics using neural temporal point processes
• Optimize hyperparameters and training lengths for temporal dependencies
• Demonstrate zero-shot predictive capability on heartbeat data

Innovation

• Density-based neural TPPs model heartbeat dynamics
• Goodness-of-fit framework optimizes hyperparameters and training lengths
• Zero-shot predictive capability demonstrated on heartbeat data