Numerical Prelim
Numerical Computation Preliminary Exam
The numerical computation is usually a one-week take home examination
consisting of a few questions requiring thoughtful
and creative answers with access to any written sources allowed.
Familiarity with the content of CSCI 5606 is assumed, including the following topics.
- Finite precision arithmetic and error: ill conditioning,
numerical stability.
- Solving systems of linear equations:
Gaussian elimination, matrix factorization, methods
for special matrices,
condition of linear equations,
error analysis, methods for large sparse systems,
nonlinear systems.
- Solving nonlinear equations:
bisection, fixed point iteration, Newton's and secant
methods, effects of rounding error,
Newton's method for systems
of equations.
- Overdetermined syetems and least squares: QR factorization, normal equations
- Interpolation and approximation:
polynomial interpolation, divided differences,
spline interpolation.
- Differentiation and integration:
trapezoidal and Simpson's rules, accuracy,
extrapolation methods.
- Methods for ordinary differential equations:
initial and boundary value problems.
- Basics of parallel computation.
Useful references are "Numerical Analysis: Mathematics of Scientific Computing"
by Kincaid and Cheney and "Scientific Computing" by Michael Heath.