WEBVTT

00:00:04.019 --> 00:00:06.768
In the next chapter about Taylor series, I make

00:00:06.767 --> 00:00:09.460
frequent reference to higher order derivatives.

00:00:10.099 --> 00:00:12.509
And if you're already comfortable with second derivatives,

00:00:12.509 --> 00:00:13.980
third derivatives, and so on, great!

00:00:14.419 --> 00:00:16.660
Feel free to just skip ahead to the main event now.

00:00:16.879 --> 00:00:17.800
You won't hurt my feelings.

00:00:18.960 --> 00:00:21.542
But somehow, I've managed not to bring up higher

00:00:21.542 --> 00:00:24.020
order derivatives at all so far in this series.

00:00:24.519 --> 00:00:26.820
So for the sake of completeness, I thought I'd give you

00:00:26.820 --> 00:00:29.079
this little footnote just to go over them very quickly.

00:00:29.640 --> 00:00:34.179
I'll focus mainly on the second derivative, showing what it looks like in the context

00:00:34.179 --> 00:00:38.560
of graphs and motion, and leave you to think about the analogies for higher orders.

00:00:40.100 --> 00:00:43.231
Given some function f of x, the derivative can be

00:00:43.231 --> 00:00:47.179
interpreted as the slope of this graph above some point, right?

00:00:47.759 --> 00:00:50.300
A steep slope means a high value for the derivative,

00:00:50.301 --> 00:00:52.460
a downward slope means a negative derivative.

00:00:53.240 --> 00:00:57.376
So the second derivative, whose notation I'll explain in just a moment,

00:00:57.375 --> 00:01:02.259
is the derivative of the derivative, meaning it tells you how that slope is changing.

00:01:03.280 --> 00:01:07.460
The way to see that at a glance is to think about how the graph of f of x curves.

00:01:08.140 --> 00:01:12.025
At points where it curves upwards, the slope is increasing,

00:01:12.025 --> 00:01:15.200
and that means the second derivative is positive.

00:01:17.799 --> 00:01:21.150
At points where it's curving downwards, the slope is decreasing,

00:01:21.150 --> 00:01:23.060
so the second derivative is negative.

00:01:26.000 --> 00:01:32.153
For example, a graph like this one has a very positive second derivative at the point 4,

00:01:32.153 --> 00:01:36.096
since the slope is rapidly increasing around that point,

00:01:36.096 --> 00:01:42.182
whereas a graph like this one still has a positive second derivative at the same point,

00:01:42.182 --> 00:01:45.640
but it's smaller, the slope only increases slowly.

00:01:46.500 --> 00:01:50.900
At points where there's not really any curvature, the second derivative is just 0.

00:01:53.379 --> 00:01:57.483
As far as notation goes, you could try writing it like this,

00:01:57.483 --> 00:02:01.318
indicating some small change to the derivative function,

00:02:01.319 --> 00:02:06.433
divided by some small change to x, where as always the use of this letter d

00:02:06.433 --> 00:02:12.085
suggests that what you really want to consider is what this ratio approaches as dx,

00:02:12.085 --> 00:02:14.439
both dx's in this case, approach 0.

00:02:15.539 --> 00:02:19.253
That's pretty awkward and clunky, so the standard is

00:02:19.253 --> 00:02:23.179
to abbreviate this as d squared f divided by dx squared.

00:02:24.360 --> 00:02:28.504
And even though it's not terribly important for getting an intuition for the second

00:02:28.503 --> 00:02:32.500
derivative, I think it might be worth showing you how you can read this notation.

00:02:33.159 --> 00:02:36.414
To start off, think of some input to your function,

00:02:36.414 --> 00:02:40.859
and then take two small steps to the right, each one with a size of dx.

00:02:42.000 --> 00:02:45.813
I'm choosing rather big steps here so we'll be able to see what's going on,

00:02:45.813 --> 00:02:49.680
but in principle keep in the back of your mind that dx should be rather tiny.

00:02:50.900 --> 00:02:55.747
The first step causes some change to the function, which I'll call df1,

00:02:55.747 --> 00:03:01.133
and the second step causes some similar but possibly slightly different change,

00:03:01.133 --> 00:03:02.480
which I'll call df2.

00:03:03.330 --> 00:03:08.990
The difference between these changes, the change in how the function changes,

00:03:08.990 --> 00:03:10.659
is what we'll call ddf.

00:03:12.020 --> 00:03:17.460
You should think of this as really small, typically proportional to the size of dx2.

00:03:18.400 --> 00:03:23.093
So if, for example, you substituted in 0.01 for dx,

00:03:23.092 --> 00:03:28.599
you would expect this ddf to be about proportional to 0.0001.

00:03:29.699 --> 00:03:34.445
The second derivative is the size of this change to the change,

00:03:34.445 --> 00:03:37.931
divided by the size of dx2, or more precisely,

00:03:37.931 --> 00:03:41.640
whatever that ratio approaches as dx approaches 0.

00:03:43.000 --> 00:03:45.955
eleration. Given some movement along a line, suppose you have some function that

00:03:45.955 --> 00:03:48.728
records the distance traveled versus time, maybe its graph looks like this,

00:03:48.729 --> 00:03:51.576
steadily increasing over time. Then its derivative tells you velocity at each

00:03:51.575 --> 00:03:53.911
point in time, for example the graph might look like this bump,

00:03:53.911 --> 00:03:56.100
increasing up to some maximum, and decreasing back to zero.

00:03:56.100 --> 00:03:57.780
So the second derivative tells you the rate of

00:03:59.039 --> 00:04:01.798
Maybe the most visceral understanding of the second

00:04:01.799 --> 00:04:04.240
derivative is that it represents acceleration.

00:04:05.180 --> 00:04:08.897
Given some movement along a line, suppose you have some function

00:04:08.897 --> 00:04:11.644
that records the distance traveled versus time,

00:04:11.644 --> 00:04:15.819
maybe its graph looks something like this, steadily increasing over time.

00:04:16.740 --> 00:04:20.225
Then its derivative tells you velocity at each point in time,

00:04:20.225 --> 00:04:24.725
for example the graph might look like this bump, increasing up to some maximum,

00:04:24.725 --> 00:04:26.300
and decreasing back to zero.

00:04:27.199 --> 00:04:29.474
The third derivative, and this is not a joke, is called jerk. So if the jerk is not zero,

00:04:29.475 --> 00:04:31.169
it means that the strength of the acceleration itself is changing.

00:04:31.168 --> 00:04:33.293
One of the most useful things about higher order derivatives is how they help us in

00:04:33.293 --> 00:04:33.900
approximating functions,

00:04:34.920 --> 00:04:39.468
In this example, the second derivative is positive for the first half of the journey,

00:04:39.468 --> 00:04:43.487
which indicates speeding up, that's the sensation of being pushed back into

00:04:43.487 --> 00:04:46.819
your car seat, or rather, having the car seat push you forward.

00:04:47.540 --> 00:04:52.520
A negative second derivative indicates slowing down, negative acceleration.

00:04:54.000 --> 00:04:57.079
The third derivative, and this is not a joke, is called jerk.

00:04:57.839 --> 00:05:03.919
So if the jerk is not zero, it means the strength of the acceleration itself is changing.

00:05:06.279 --> 00:05:09.655
One of the most useful things about higher order derivatives is

00:05:09.656 --> 00:05:13.138
how they help us in approximating functions, which is exactly the

00:05:13.137 --> 00:05:16.620
topic of the next chapter on Taylor series, so I'll see you there.