Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 27, 2026Digital Government Research and PracticeOpen Access

HiDiNet: High-Dimensional Interpretive Network for Modeling Aging Health and Survival

View Full Paper
Ask AI
Bookmark
Share

Authors

HGHannah GuanADAashish DhananiYZYu Zhang

Discussion

Loading...

Member takes

Overview

Randomized trial reveals improved health trajectory predictions in aging populations, suggesting powerful applications in healthcare.

Key Points

  • This research aims to model health trajectories and survival in aging individuals using a new predictive framework.
  • Developed HiDiNet to use stochastic differential equations for health evolution.
  • Integrated Transformer-based attention for capturing long-range temporal dependencies.
  • Evaluated on the English Longitudinal Study of Ageing with 10 waves from 1998 to 2019.
  • HiDiNet achieved a Brier score of 0.33, outperforming RNN (0.42) and Elastic-Net (0.70).
  • C-index for HiDiNet was 0.968 compared to 0.951 for RNN and 0.700 for Elastic-Net.
  • D-calibration showed HiDiNet's reliability (p = 0.932) versus 0.280 for the Transformer-only model.

Cite This Study

Guan et al. (2026) studied this question.

synapsesocial.com/papers/6a6700af40bca442e0d4a93ehttps://doi.org/10.1145/3833874
View Full Paper
Ask AI
Bookmark
Share