Numerical Diff of Continuous Functions - First Derivative: Forward Divided Diff
|
In this segment we'll talk about
numerical differentiation of continuous functions. We'll talk about how to
find the first derivative, and the method which we are considering is the Forward
Divided Difference method. So, if you remember your differential calculus, you
must have come across this this expression here that if you want to find the
derivative at a certain value of x then f(x), f'(x) is simply defined by the
limit of let's suppose h approaching 0 of this expression right here. So,
what the Forward Divided Difference method does is simply says that hey I
cannot choose h approaching 0 so let me choose h approaching a finite number.
So, what we do is we take a point let's suppose x + h here so the distance
between these two points is h. And what we do is we draw a secant line
between these two points. So, since that's the secant line and this
coordinate will be given by this. This coordinate here is this. And if we
tend to find out what the slope of the secant line is, that'll be this rise
over this run, so we already know the run is h, what is the rise? Rise is the
difference between the function at x + h and the function at x. So, what Forward
Divided Difference method is simply taking the rise over run. That's about it.
So, the difference between the exact definition and the Forward Divided Difference
definition is just that h is a finite number. So that is the Forward Divided
Difference method let's go and take an example and see how we can find this approximate
value of the derivative and what this comparison will be with the true value and also see therefore the effect of the step size h here
is on the finding the derivative of the function. So, this is an example we are
taking. We are asked to find the derivative of this function at 3, with a
step size of 0.16, we’re supposed to find the approximate value, we're
supposed to find the true value, which is from your differential calculus
class, the true error will be the difference between the true value and the
approximate value. We're also supposed to discuss the trends in the true
error as a function of the step size, and also show
that hey how is the absolute relative approximate error, which you're going
to calculate in this case related to the significant digits which are correct
as the step size is decreased. So, let's go ahead and solve each of these
parts of the problem, one by one. So, let's do part a. We already
know that the approximate value of the derivative function is given by this
formula by the Forward Divided Difference method. Our x is 3, the step size
which was supposed to take is 0.16, the function is given to us 7x^4. So, in order to find the value of the function at the
derivative of the function of 3, it'll be substituting the value of x and h
here, and now what we're going to do is, so this becomes the value of the
function at 3.16 minus the value of the function of 3 divided by 0.16. And
what is the value of the function at 3.16? It will be 7*x to the power 4, and
that simply gives us a value of 818.6591. So that is the value approximate
value of the derivative of the function x = 3. Let's go and see what is the. So,
if we want to find the true value f of x is given as 7 x to the power 4, f
prime of x will be given by 7 times 4 x to the power 3, that's the derivative
x to the power 4 is 4 x cubed, the 28 x cubed. Hey
what is the value of the derivative of the function at 3, it is 28 times 3
cubed, and that turns out to be exact value of 756. So, if that is the exact
value, what is the true error? The true error will be true value minus
approximate value. The true value we got was 756, approximate value which we
got 818.6591. And that gives us a true error of the following number which is
-62.6591. So, if you look at in relationship to the true value this true
error is about eight percent error, or so. The next one is asking us that,
hey, can you show us what happens to true errors as the step size is decreased? So, I can repeat the process, but I would
like you to do it. So, step size is h, let's suppose, and we are asked to
find the approximate derivative of f prime of 3
then we are supposed to look at the true error. We just found out that for
0.16, f’(3) is 818.6591. The true error is as follows. Then what I’ll do is,
let's suppose, if I have the step size and you repeat the same process as I
did before, we get this as our approximate value, and the true error is this
much. If we again have the step size, the approximate value which I get is
this, and the true error is now something like this. So, what you're finding
out here is as the step size is decreased the approximate value starts to
become closer and closer to the exact value, which is 756. And
also, as we're calculating our true error which is the difference
between the exact value and the approximate value, you find out that the
trend of the true error magnitude is also decreasing. But one of the things
which I would like you to observe here is that what happens to the true error
is it getting approximately halved. See about half of 62 is 31, half of 30 is
15, so it's getting approximately half. And this is something which you
should recognize because later on we'll talk about
what is the how does the error change as a function of step size there's a
separate lesson, but just an illustration here tells you that hey we are
feeling that the true error is getting is proportional to the step size as I
have the step size the true error is getting approximately halved. Let's look at the last part
where we are asked to say hey, can you tell us what can what happens to the
relative approximate error? So, in this case again although we have already
shown it for h = 0.16, you can do it for other step sizes as well which is
the repetition of the same process. So, we're finding f prime of 3, we will
find what the approximate error is, and then we're asked to find the absolute
relative approximate error. So, we already know that with a value of h =
0.16, we got f prime of 3 to be this quantity here. And so, the approximation
cannot be calculated at this point because we don't have a previous
approximation. So, we'll keep those blank. Now when I go to 0.08 step size, this
is the value which you'll get if you repeat the process as we did this in
part a. Then what is approximate error? Approximate error is the difference
between this value minus this value. Current approximation minus the previous
approximation. And that turns out to be this quantity right here. The relative
approximate error will be the ratio between the approximate error and the
current approximation. And that turns out to be 4.0517%. If we have a step
size of 0.04, and i repeat the process as in part a,
this is what I will get as my value of the function derivative at 3. The
approximate error will be the difference between the current approximation
and the previous approximation which turns out to be 15.5263. And the
absolute relative approximate error turns out to be 2.0131%. And the reason
why we talk about this or why we have to discuss
this is because in real life we won't be privy to the exact value of the
derivative of the function. So, we'll have to make our judgment call of the
errors based on the absolute relative approximate error. So, for example in
this case it turns out to 2.013%. So, we know that 2.013% is less than or
equal to five percent, but not less than or equal to 0.5%. So, what that
implies is that hey at least one significant digit is correct in our answer.
So, when you look at 771.2548 as the answer for our h = 0.04, that’s our f
prime of 3, we can trust this significant digit, which is the 7 right here,
in that answer. And that is the end of this segment. |