The Non-Linear Equations slides
can be found in the
common
directory:
NL-eqn.pdf.
Do the following problems
From our text
Problems P2.21
Extra Problems:
We discussed in class that if
-LOG10(|x-x*|/
|x*|) = k
and x is close to x*,
the approximation value x and the true value
x* have about k most
significant digits identical.
Prove this claim with a convincing argument.
Do keep in mind
that prove-by-example is NOT a proof.
Prove-by-example
means the use of an example to prove
a proposition.
Write a small Fortran program to compute the
variance of three values with the one-pass and
two-pass methods. Then, run it with 9000000,
9000001 and 9000002 as input. Is the computed
variance 1?
Use hand-calculation without the help of a calculator
to determine what went wrong in your program's
output.
Use fixed-point iteration to find
a root of
f(x) = e-x
- x2.
Do the following:
Use
gnuplot to plot the graph of
f(x) = e-x
- x2 to locate a root.
Transform the equation to a form of
x = g(x).
Use
gnuplot to plot a graph of
x = g(x)
and y = x, and check the
tangent line slop of
x = g(x) to ensure
fixed-point iteration can converge.
Use your calculator to do a fixed-point
iteration.
Note that
you should avoid all kinds of floating point
problems (e.g., over- and under- flow,
division-by-0, cancellation, etc)
.
Otherwise, you should try a different
transformation and/or a new initial guess.
You do not have to turn in your paper. What I really expect
you to do is using these problems to gauge your understanding
of the subject. So, do the problems after finish
reading the above sections.