Last time: | Next time: |
Today:
You might read through 8.5 for next (RK methods).
I may be making a few references to that page as we go along. Let's start with this one (about what we're actually solving, or attempting to solve).
The upshot is expressed well in the authors' quote that "...the numerical solution values and do not lie on the true solution. In fact, they do not even lie on the same solution."
Start by taking a peak at Figure 8.13, p. 325. It illustrates everything that we want to consider.
Our author distinguishes three types of error: in this figure there are two,
I want to prove a result which gives a bound on the error we're making (ignoring rounding error):
We could use this formula to find a maximum step-size to take to control for a given global error , by bounding the total error at the right endpoint, :
Compare this to the result at the bottom of p. 326. They claim that the global discretization error is proportional to .
If you expand the exponential in a Taylor series, you'll get
A rough guide, then, for a step-size would be
The neat thing is that we can arrive at an optimal to reduce the error at the end of interval, when ,
provided we can bound the second derivative and the rounding errors:
Questions:
What we notice in the Rabbits and Foxes example is that Euler doesn't do a very good job. In the phase plane, the cycle is not closed (whereas it appears to be for NDSolve -- so maybe Mathematica's routine is better than Euler!:).
Let's look at an example: the harmonic oscillator (a simplification of the simple pendulum).
In particular, we generalize the formula on p. 323,
to the more general formula
Note the parameter in the function: will generally be a function of .
And so on! That's the idea of the higher-order Taylor methods.
Let's have a look at a derivation with a slightly different emphasis from that in our text (our authors want to avoid partial derivatives, but I don't!).
Examples: