HyperNSDE: Personalized Neural SDEs for Joint Static-Longitudinal Clinical Data Generation
Abstract
Synthetic patient data generation is a promising solution to the dual challenge of data scarcity and privacy constraints in healthcare machine learning. Realistic synthesis of patient-level clinical data requires jointly modeling heterogeneous static covariates, irregularly sampled longitudinal trajectories, and informative observation times - three tightly coupled components in practice yet rarely addressed together. We propose HyperNSDE, a continuous-time generative model that conditions a latent Neural SDE on static patient representations through a hypernetwork, allowing baseline characteristics to shape trajectory evolution beyond the initial condition without requiring a trajectory encoder, while stochastic latent dynamics capture realistic variability in generated paths. Observation times are modeled jointly through a latent-state-dependent intensity process, and training on irregular stochastic paths is stabilized via a deterministic-stochastic path decomposition with a non-adversarial signature-kernel objective. Experiments on simulated and real clinical datasets show improved observation-time fidelity and competitive performance, while matched-grid analyses reveal that forecasting and correlation metrics are affected by observation-grid regularity and trajectory smoothness.
Related
- DagFlow AI: Agentic Synthetic Data Generation for Personalized Statistics Learning
- Latent Similarity Gaussian Processes: A Theory-Grounded Approach to Personalized Suicide-Risk Forecasting for Clinical Decision-Support
- Bridging Denotator Theory and Sheaf Neural Networks for Symbolic Music Generation
- Stochastic dynamics generation of directed percolation using a two-channel convolutional neural network
- Conditional Neural SDEs for Financial Time-Series Generation
BibTeX
@misc{chassat2026hypernsde,
title = {{HyperNSDE: Personalized Neural SDEs for Joint Static-Longitudinal Clinical Data Generation}},
author = {Perrine Chassat and Agathe Guilloux},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.07383},
eprint = {2610.07383},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.07383},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - HyperNSDE: Personalized Neural SDEs for Joint Static-Longitudinal Clinical Data Generation
AU - Perrine Chassat
AU - Agathe Guilloux
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.07383
UR - https://arxiv.org/abs/2610.07383
AB - Synthetic patient data generation is a promising solution to the dual challenge of data scarcity and privacy constraints in healthcare machine learning. Realistic synthesis of patient-level clinical data requires jointly modeling heterogeneous static covariates, irregularly sampled longitudinal trajectories, and informative observation times - three tightly coupled components in practice yet rarely addressed together. We propose HyperNSDE, a continuous-time generative model that conditions a latent Neural SDE on static patient representations through a hypernetwork, allowing baseline characteristics to shape trajectory evolution beyond the initial condition without requiring a trajectory encoder, while stochastic latent dynamics capture realistic variability in generated paths. Observation times are modeled jointly through a latent-state-dependent intensity process, and training on irregular stochastic paths is stabilized via a deterministic-stochastic path decomposition with a non-adversarial signature-kernel objective. Experiments on simulated and real clinical datasets show improved observation-time fidelity and competitive performance, while matched-grid analyses reveal that forecasting and correlation metrics are affected by observation-grid regularity and trajectory smoothness.
ER -