CAS Seminar
Marcel Oliver
(Jacobs University)
"A dynamic moving least square algorithm"
| Date: |
Tue, November 19, 2019 |
| Time: |
14:15 |
| Place: |
Research I, Room 103 |
Abstract: Dynamic moving least squares may provide a systematic and rational way to
- provide interpolation, thus differentiation, on structured and unstructured meshes alike,
- allow for the presence of noise (either from the data, or from accumulated errors in the time integration algorithm),
- adapt the local order or accuracy to the data,
- adapt the local stencil to the data.
I will report on a Bachelor-thesis project from a few years ago by Dan Erdmann-Pham who designed a prototype of this algorithm which appears to be new, generalizes and unifies the concepts of (W)ENO methods on the one hand and MLS on the other hand, and is based on a fixed-point concept that induces a local-global interaction necessary to satisfy all the above requirements.
