CADENCE: Recovering Individual Trajectories from Sparse Snapshots
A new probabilistic framework called CADENCE recovers continuous individual trajectories from isolated cross-sectional snapshots, challenging the traditional need for dense longitudinal data. Developed by researchers whose names are not provided in the source, CADENCE anchors latent dynamics to static, individual-level contexts, enabling inference from extremely sparse or single-timepoint data. The framework offers novel identifiability guarantees for single-timepoint trajectory inference, addressing a fundamental ill-posed problem in fields such as aging, epidemiology, and physical degradation. Unlike sequence models (e.g., latent ODEs) that require dense longitudinal data or cross-sectional methods (e.g., optimal transport, flow matching) that lose individual dynamics, CADENCE bridges the gap by combining a score-based approach with principled probabilistic modeling. The paper is published on arXiv under identifier 2605.23470v1.
Key facts
- CADENCE is a probabilistic framework for recovering individual trajectories from sparse cross-sectional snapshots.
- It anchors latent dynamics to static, individual-level contexts.
- It provides identifiability guarantees for single-timepoint trajectory inference.
- The framework addresses ill-posed problems in aging, epidemics, and physical degradation.
- It combines a score-based approach with probabilistic modeling.
- The paper is on arXiv with identifier 2605.23470v1.
- Sequence models like latent ODEs require dense longitudinal data.
- Cross-sectional methods like optimal transport lose individual dynamics.
Entities
Institutions
- arXiv