Search papers, labs, and topics across Lattice.
This paper introduces a novel Finsler metric for trajectory inference that combines continuous geometric priors with discrete, directed prior knowledge, such as lineage trees. By incorporating both geometric and classification-based priors, the method learns lineage-guided geodesics that better reflect the underlying dynamics of the system. Experiments on synthetic and real-world data demonstrate improved performance on trajectory interpolation tasks compared to methods relying solely on Riemannian metrics.
By blending geometry with classification, this new Finsler metric lets you trace trajectories more accurately through complex systems, like cell development, where you have both spatial data and lineage trees.
Trajectory inference investigates how to interpolate paths between observed timepoints of dynamical systems, such as temporally resolved population distributions, with the goal of inferring trajectories at unseen times and better understanding system dynamics. Previous work has focused on continuous geometric priors, utilizing data-dependent spatial features to define a Riemannian metric. In many applications, there exists discrete, directed prior knowledge over admissible transitions (e.g. lineage trees in developmental biology). We introduce a Finsler metric that combines geometry with classification and incorporate both types of priors in trajectory inference, yielding improved performance on interpolation tasks in synthetic and real-world data.
