Research Program
Dynamic Reconstruction
Dynamic Reconstruction studies how a time-dependent cognitive state can be inferred from incomplete multimodal measurements. The program separates structural constraints from transient activity and estimates the state trajectories most consistent with observed neural, behavioral, linguistic, and affective signals.
Program Code
NSR-17
Status
Active
Lead Domain
Computational Neuroscience
Summary
Dynamic Reconstruction studies how a time-dependent cognitive state can be inferred from incomplete multimodal measurements. The program separates structural constraints from transient activity and estimates the state trajectories most consistent with observed neural, behavioral, linguistic, and affective signals.
Method
The program combines recurrent state-space models, longitudinal acquisition, held-out challenge tasks, and uncertainty-aware reconstruction. Models are evaluated on their ability to reproduce response dynamics rather than static similarity alone.
Current Objective
Reduce reconstruction error under sparse or noisy acquisition while preserving temporal correspondence.
Related Publications
- EXIST Technical Paper 16.002
- Continuity Systems Paper 21.011