Block course (LSF)
- Dr. Wayne Mitchell
- Time, 9:15 - 15:45
- No exam, no ECTS
- Heidelberg, INF 350 / OMZ R U014
- Mo 17.02.2020 - Fr 21.02.2020
Contents
- Day 1: Basic iterative methods, Krylov methods and preconditioning
- Day 2: Geometric multigrid
- Day 3: Algebraic multigrid
- Day 4: Special topics: non-linear problems, non-symmetric problems, and graph problems
- Day 5: High-performance parallel AMG and communication-avoiding multigrid algorithms
Description
This short, intensive course will introduce students to foundational concepts as well as current research topics in multigrid methods for solving large-scale linear and non-linear systems of equations.
The course will begin by introducing basic iterative methods and motivating the underlying ideas of multigrid methods. The course will then proceed to cover both geometric and algebraic multigrid methods. Some special cases and extensions to non-linear and non-symmetric problems will also be discussed. Finally, special considerations for parallel implementations of multigrid algorithms will be covered along with new and ongoing research on communication-avoiding variants.
Students are assumed to have basic background in linear algebra and have at least some familiarity with partial differential equations, discretizations, and scientific computing.
Morning sessions will consist of two lecture blocks, each approximately 1.5 hours, and the afternoon will be reserved for hands-on exercises, questions, and discussion.
Because there is no exam at the end, one can also attend only part of the course, e.g. only attend the first introductory sessions, or if you are already familiar with the basics of multigrid, only attend the later more advanced sessions. No registration required.