Tiejun Li

We address the multiscale singularity in Euclidean diffusion models for manifold-constrained data and propose two novel methods to improve sampling accuracy on complex geometries.

Var-RUOT is a new framework improving RUOT for recovering dynamics from high-dimensional snapshots. It incorporates first-order optimality conditions, achieving better performance on single-cell datasets.

We introduce CytoBridge, a new deep learning approach that models unbalanced stochastic cellular dynamics and interactions from snapshot data.