2

Gaussian Process Subspace Regression for Model Reduction

A fast and accurate method that approximates subspace-valued functions, with exampmle applications in projection-based parametric reduced order modeling.

Newton Retraction as Approximate Geodesics on Submanifolds

Newton retraction is always stabler and faster than orthographic retraction, aka oblique projection.

Normal-bundle Bootstrap

A method that generates new data which preserve the geometric structure of a given data set, with applications in inference and data augmentation.

Drivers Learn City-scale Dynamic Equilibrium

Simplistic, testable models of collective human behavior and dynamics can emerge from individual-level mechanisms that correctly capture incentives.

Demand, Supply and Performance of Street-Hail Taxi

Taxis out-perform TNCs such as Uber in high-demand locations, which suggests a taxi/TNC regulation change to reduce congestion and pollution.