43 min
Seminar, Hosted by Simons Institute for the Theory of Computing
A new class of algorithms for trajectory inference
Yale University
Recording
Abstract
Trajectory inference reconstructs smooth paths from unpaired observations collected at different times. Acceleration Matching lifts the interpolation problem to phase space and learns an explicit conditional acceleration field. The resulting method uses positional observations without costly preprocessing or simulated training trajectories, with numerical comparisons against established approaches.
Topics
- Trajectory inference, broader topic: inference
- Acceleration Matching, search World Wide
- phase space
- conditional acceleration field, search World Wide
- interpolation
- positional observations, broader topic: observations
- simulated training trajectories, search World Wide
- numerical comparisons, search World Wide
- unpaired observations, broader topic: observations
- smooth paths, search World Wide