Trajectory Inference in Wasserstein Space

Learn about trajectory inference methods in Wasserstein space through this 51-minute research talk that explores B-spline approximation and interpolation techniques for reconstructing continuous processes from cross-sectional measurements. Dive into innovative approaches combining subdivision schemes with optimal transport-based geodesics to handle dynamic processes in computational biology. Discover how these methods achieve trajectory inference with customizable precision and smoothness, particularly useful for scenarios involving particle division over time. Examine the proven linear convergence rates and rigorous evaluation results on cell data featuring complex scenarios like bifurcations, merges, and trajectory splitting in supercells. Compare the performance against current trajectory inference and interpolation methods in this collaborative research presentation by Caroline Moosmüller, developed with Amartya Banerjee, Harlin Lee, and Nir Sharon.