MTH 655
and
MTH 659 (Numerical Analysis)
Large scale scientific computing methods
- Winter 2013
|
|
| General information |
|---|
INSTRUCTOR:
Malgorzata Peszynska
CLASS:
MWF 9:00-9:50 Rogers 440
COURSE INFORMATION:
In this class we develop theory and implementation details for solving
large scale scientific computing problems. - Rigorous
mathematical background as well as algorithms and implementation
details will be developed for solving large linear and nonlinear
systems of equations using Newton-Krylov methods, multigrid and domain
decomposition. These arise typically from discretizations of
(continuum) partial differential equation models.
- We will also develop background and applications for discrete
(lattice) and Monte Carlo methods. These are bread-and-butter
computational physics models and are easily parallelizable.
- Simple model case studies as well as examples from applications
will be developed in which scientific computing techniques that you
will learn in this class will be applied. Many of these have
non-textbook properties or structure. For example, semismooth Newton
methods are helpful in non-differentiable models.
Students will be introduced to parallel computing and will learn how
to function in a high performance computing environment.
STUDENTS: The course is intended for graduate students
of mathematics and other disciplines and for well-prepared
undergraduates. No specific preparation beyond solid undergraduate
background in mathematics will be assumed. Knowledge of numerical
methods, and familiarity with computer programming are a plus but are
not required: students will be graded based on their learning
derivative.
Students are encouraged to contact
me with questions about
the class.
SEQUENCE MTH 654-656 in 2012-2013:
The courses in this sequence can be taken independently and
are taught by different instructors.
GRADING:
- Attendance at all labs Fridays in MLC Kidd 108J
is required. Please contact me if you have to miss a lab meeting.
You have to complete all lab projects and turn in lab reports.
- The class will emphasize problem solving, breadth and creativity.
Quality of the work and the derivative of your learning curve rather
than the absolute measure of performance will determine your grade.
Course Outcomes: A successful student will be able to
- Understand, analyze, and implement basic iterative methods
for nonlinear and linear equations in N dimensions
- Determine convergence of an algorithm theoretically and experimentally
- Implement basic algorithms in interpretive computational
environments such as MATLAB as well as in traditional scientific
computing environments on remote platforms using compilers and
libraries
- Understand the basic principles of parallel computing and domain decomposition
- Solve selected applications problems arising from continuum and
discrete mathematical modeling and computational physics
Special arrangements for students with disabilities, make-up
exams etc.: please contact the instructor and Services for Students with
Disabilities, if relevant, and provide appropriate documentation.
Course drop/add information is at
http://oregonstate.edu/registrar/.
|
|
|
