Useful info
Course website: http://math.gmu.edu/~memelian/teaching/Spring10 MATLAB instructions: http://math.gmu.edu/introtomatlab.htm Mathworks, the creator of MATLAB: http://www.mathworks.com OCTAVE = free MATLAB clone Available for download at http://octave.sourceforge.net/
Scientific computing
Design and analysis of algorithms for numerically solving mathematical problems in science and engineering Deals with continuous quantities vs. discrete (as, say, computer science) Considers the effect of approximations and performs error analysis Is ubiquitous in modern simulations and algorithms modeling natural phenomena and in engineering applications
Develop mathematical model (usually requires a combination of math skills and some a priori knowledge of the system) Come up with numerical algorithm (numerical analysis skills) Implement the algorithm (software skills)
Mathematical modeling
Run, debug, test the software Visualize the results Interpret and validate the results
The problem is well-posed, if (a) solution exists (b) it is unique (c) it depends continuously on problem data The problem can be well-posed, but still sensitive to perturbations. The algorithm should attempt to simplify the problem, but not make sensitivity worse than it already is. Simplification strategies: Infinite finite Nonlinear linear High-order low-order
Before computation modeling approximations Cannot be controlled empirical measurements, human errors previous computations During computation Can be controlled through truncation or discretization error analysis Rounding errors Accuracy depends on both, but we can only control the second part Uncertainty in input may be amplified by problem Perturbations during computation may be amplified by algorithm Abs_error = approx_value true_value Rel_error = abs_error/true_value Approx_value = (true_value)x(1+rel_error)
G(y) F(x)
Total error
=
=
[G(y) - F(y)]
[F(y) - F(x)]
Computational error: Propagated data error: + affected by algorithm not affected by algorithm
Rounding error: introduced by finite precision calculations in the computer arithmetic Truncation error: introduced by algorithm via problem simplification, e.g. series truncation, iterative process truncation etc. Computational error = Truncation error + rounding error
How much must original problem change to give result actually obtained?
How much data error in input would explain all error in computed result?
Approximate solution is good if it is the exact solution to a nearby problem Backward error is often easier to estimate than forward error
Conditioning
Well-conditioned (insensitive) problem: small relative change in input gives commensurate relative change in the solution Ill-conditioned (sensitive): relative change in output is much larger than that in the input data Condition number = measure of sensitivity
Condition number = |rel. forward error| / |rel. backward error| = amplification factor
Conditioning
Stability
Algorithm is stable if result produced is relatively insensitive to perturbations during computation Stability of algorithms is analogous to conditioning of problems From point of view of backward error analysis, algorithm is stable if result produced is exact solution to nearby problem For stable algorithm, effect of computational error is no worse than effect of small data error in input
Accuracy
Normalized representation
Not all numbers can be represented this way, those that can are called machine numbers
Rounding rules
If real number x is not exactly representable, then it is approximated by nearby floating-point number fl(x)
This process is called rounding, and error introduced is called rounding error
Two commonly used rounding rules
chop: truncate base- expansion of x after (p 1)st digit; also called round toward zero round to nearest : fl(x) is nearest floating-point number to x, using floatingpoint number whose last stored digit is even in case of tie; also called round to even
Round to nearest is most accurate, and is default rounding rule in IEEE systems
Machine precision
Loss of significance
Loss of significance
Loss of significance
Loss of significance