Abstract: In the Multivariate Functional Approximation (MFA) project, B-splines are used to approximate discrete scientific data. A B-spline is a piecewise polynomial curve defined by its degree, knot vector, and control points. Given a degree and knot vector, a set of control points can be solved to approximate the input data to some accuracy. However, the choice of knot vector has immense influence in the resulting accuracy of the approximation.
In this talk, I will share a fast and automatic method I developed to determine knot vectors for accurate B-spline approximation. The method uses features generated from high-order derivatives of the data points. I will compare the approximations resulting from my approach with state-of-the-art methods.
Bio: Raine Yeh is a Ph.D. student in computer science at Purdue University. Her research is on data analytics and management of scientific data in high-performance computing.