講 題：Hessian estimators on data manifolds
Given a function defined on a data set, there have been several ways to estimate the Hessian of it. Although the estimators can be quite accurate, their constructions are all based on certain grid structures. Without using any grid assumption, we construct a Hessian estimator which converges to the continuous Hessian tensor when the data points are densely sampled from a data manifold. This is a joint work with Dr. Hau-Tieng Wu.