## Help with the Chain Rule

For the discussion of math. Duh.

Moderators: gmalivuk, Moderators General, Prelates

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Help with the Chain Rule

I was recently taught about the Chain Rule in calculus class, and I am really confused. My biggest problem is:

(d/dx) f(x) = f'(x) And
(d/dx) f(u) = f'(u) * u' Substitute (d/dx) f(u) with f'(u) and you get
f'(u) = f'(u) * u'

While I am asking questions, how is a corner/ cusp different than a curve, and what is a tangent line. I was told that it is a line that touches a graph (locally) at only one point, but then a vertical line is the tangent of any point of a function by definition of a function.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

>-)
Posts: 527
Joined: Tue Apr 24, 2012 1:10 am UTC

### Re: Help with the Chain Rule

the second equation is (without using prime notation)
(d/dx) f(u) = (d/du) f(u) * (d/dx) u
and it is not the case that (d/dx) f(u) = (d/du) f(u) = f'(u) so the substitution does not work

intuitively, the tangent line lies "parallel" to the curve at the point it is touching, although wikipedia gives a more rigorous definition

Xanthir
My HERO!!!
Posts: 5426
Joined: Tue Feb 20, 2007 12:49 am UTC
Contact:

### Re: Help with the Chain Rule

For another intuition, a tangent touches a curve in exactly one point but does not intersect it. (At least locally - "far" from the tangent point, the line might intersect the curve, but we're typically just interested in the "local" behavior when talking about tangents.)

When talking about tangent lines in calculus, tho, that "it's parallel to the point it's touching" is the intuition you typically want - it's "what would happen if the curve just kept going straight at this point, rather than continuing to curve?".
(defun fibs (n &optional (a 1) (b 1)) (take n (unfold '+ a b)))

Sizik
Posts: 1261
Joined: Wed Aug 27, 2008 3:48 am UTC

### Re: Help with the Chain Rule

Xanthir wrote:For another intuition, a tangent touches a curve in exactly one point but does not intersect it. (At least locally - "far" from the tangent point, the line might intersect the curve, but we're typically just interested in the "local" behavior when talking about tangents.)

Doesn't quite work for example with y = x3 at the point (0, 0), where the tangent line y = 0 does in fact intersect the curve at that point.
she/they
gmalivuk wrote:
King Author wrote:If space (rather, distance) is an illusion, it'd be possible for one meta-me to experience both body's sensory inputs.
Yes. And if wishes were horses, wishing wells would fill up very quickly with drowned horses.

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

Yeah, the "not intersecting" specification should not be considered part of the definition of tangency at all.
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

Xanthir
My HERO!!!
Posts: 5426
Joined: Tue Feb 20, 2007 12:49 am UTC
Contact:

### Re: Help with the Chain Rule

It's not, it's part of an intuition about tangency, to distinguish the tangent from all the other lines that just intersect the curve at one point.

It's definitely not exact.
(defun fibs (n &optional (a 1) (b 1)) (take n (unfold '+ a b)))

cyanyoshi
Posts: 419
Joined: Thu Sep 23, 2010 3:30 am UTC

### Re: Help with the Chain Rule

A tangent line is simply the line that best approximates the curve when you zoom in reeeeally close to a given point. Lines are generally much easier to work with than curves, so looking at a tangent line instead can give you a good sense of what the curve is doing, when the curve itself may be too complicated to work with. This kind of thinking happens all the time in engineering.

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Re: Help with the Chain Rule

>-) wrote:the second equation is (without using prime notation)
(d/dx) f(u) = (d/du) f(u) * (d/dx) u
and it is not the case that (d/dx) f(u) = (d/du) f(u) = f'(u) so the substitution does not work

You are taking the derivative of a function with respect to another function. What does that mean? Also, if we are allowed to take derivatives with respect to functions, then what does the derivative of a function with respect to itself equal?

P.S. I am fairly certain of this, but I just want to check; when we write u, we actually mean u(x).

Wikipedia wrote:More precisely, a straight line is said to be a tangent of a curve y = f (x) at a point x = c on the curve if the line passes through the point (c, f (c)) on the curve and has slope f '(c) where f ' is the derivative of f.

My teacher said that the only time a function has a limit, and is not differentiable is when there is a corner, cusp, or a vertical tangent. Therefor, Wikipedia's definition of a tangent is flawed.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

jewish_scientist wrote:You are taking the derivative of a function with respect to another function. What does that mean? Also, if we are allowed to take derivatives with respect to functions, then what does the derivative of a function with respect to itself equal?

P.S. I am fairly certain of this, but I just want to check; when we write u, we actually mean u(x).

The limit definition you learned stays good, as does the general intuition for da/db as the ratio of the change in a to the change in b when the change in b gets very small. And the ratio of the change in u to the change in u is always 1 (including when the change in u is small.)

My teacher said that the only time a function has a limit, and is not differentiable is when there is a corner, cusp, or a vertical tangent. Therefor, Wikipedia's definition of a tangent is flawed.

The functions you are dealing with so far are generally of the form y(x). There, there is something special going on with verticalness. Wikipedia is using a notion of tangency which is appropriate to math more advanced than calc 1. It just gets better and better! If you think of a function parametrized like (x(t),y(t)), it will be very comfortable indeed to talk about vertical lines as tangents.
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

ConMan
Shepherd's Pie?
Posts: 1691
Joined: Tue Jan 01, 2008 11:56 am UTC
Location: Beacon Alpha

### Re: Help with the Chain Rule

What's the volume of a sphere, as a function of its radius? V(r) = 4/3 pi r³, right? So what's the derivative of the volume with respect to the radius? That would be dV/dr = 4 pi r² (which happens to be the surface area, which is a nifty thing, but let's not dwell on that for now). So in other words, the rate at which the volume changes as the radius changes is proportional to the square of the radius itself. With me so far?

Well now, what if the radius is, itself, a function of something else? Maybe we're inflating a spherical balloon such that r(t) = sqrt(t), i.e. it increases as the square root of time. So at t = 0, r = 0, and at t = 4, r = 2, and so forth. Then dr/dt = 1/2 t-1/2, I think you'll find. Well so what? Well, how is the volume of the balloon changing as a result?

We can just substitute directly in this case, giving us V(t) = V(r(t)) = 4/3 pi t3/2 and then differentiating to get V'(t) = 2 pi t1/2. Or, we can use the chain rule. Knowing that dV/dt = dV/dr × dr/dt, we get dV/dt = 4 pi r² × 1/2 t-1/2 = 4 pi t × 1/2 t-1/2 = 2 pi t1/2.

Sometimes, substitution isn't quite so pretty, and the chain rule is the only way to make it work. And you can even use it to pull tricks like 1 = dy/dy = dy/dx × dx/dy, hence the derivative of y with respect to x is the inverse of the derivative of x with respect to y, i.e. at the equivalent points of inverse functions, the slopes of the tangents are inverses of each other.

As for more complicated stuff, check what doogly just said. Once you start introducing extra dimensions everything gets fun and you can do stuff you never thought possible.
pollywog wrote:
Wikihow wrote:* Smile a lot! Give a gay girl a knowing "Hey, I'm a lesbian too!" smile.
I want to learn this smile, perfect it, and then go around smiling at lesbians and freaking them out.

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

One day, the ultimate power of the tangent bundle may be yours!
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

Repeekthgil
Posts: 15
Joined: Wed Jan 30, 2013 6:02 pm UTC

### Re: Help with the Chain Rule

The chain rule applies for nested functions. I like to think of them as having an inner and out part. I teach my students, in addition to the notation you've written that the chain rule is the following sentence:

"the derivative of the outside, with the inside unchanged, times the derivative of the inside."

For some (most so far, but it's only my 3rd year teaching Calc) that is a helpful way to see the chain rule.

Are you learning this at college or high school? I'm asking b/c if it's high school that makes me feel good about the pacing of my class!

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Re: Help with the Chain Rule

Repeekthgil wrote:The chain rule applies for nested functions. I like to think of them as having an inner and out part. I teach my students, in addition to the notation you've written that the chain rule is the following sentence:

"the derivative of the outside, with the inside unchanged, times the derivative of the inside."

For some (most so far, but it's only my 3rd year teaching Calc) that is a helpful way to see the chain rule.

Yeah, that is how my teacher describes it. One part that I have trouble wrapping my head around is that it does not matter where the inner function is.

f(u) = stuff + u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff * u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff / u then d/dx(u) = d/du f(u) * d/dx u(x)

Are you learning this at college or high school? I'm asking b/c if it's high school that makes me feel good about the pacing of my class!

My high school (I really did not like that place) required 3 years of math. However, one of those unwritten rules of what was that all seniors take calculus or statistics. I decided to do neither. The only classes I took were studies, sciences and 1 English (it was required). I think I had 4 sciences that year. They were physics, disease diagnostics (a.k.a. House the class),Destination Imagination, and engineering. It was my best school year, so far. So I am in college now, I did have the opportunity to take calculus.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

jewish_scientist wrote:Yeah, that is how my teacher describes it. One part that I have trouble wrapping my head around is that it does not matter where the inner function is.

f(u) = stuff + u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff * u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff / u then d/dx(u) = d/du f(u) * d/dx u(x)

Actually none of these seem right. By d/dx(u), do you mean du/dx? That wouldn't involve any chain rule. If you mean something else, the notation is very weird.

If you have f(u), df/dx = df/du * du/dx.

And then it ought to make sense that the chain rule doesn't depend on what f(u) is at all. Right? What the chain rule is saying is that
df/dx = df/du * du/dx
How f(u) looks will matter an awful lot for what df/du is. If you go to actually implement these things, it makes a big difference.

But the chain rule itself is about partitioning these dependencies. It doesn't care about the actual functional form of f(u), or of u(x).
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Re: Help with the Chain Rule

doogly wrote:
jewish_scientist wrote:Yeah, that is how my teacher describes it. One part that I have trouble wrapping my head around is that it does not matter where the inner function is.

f(u) = stuff + u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff * u then d/dx(u) = d/du f(u) * d/dx u(x)
f(u) = stuff / u then d/dx(u) = d/du f(u) * d/dx u(x)

Actually none of these seem right. By d/dx(u), do you mean du/dx? That wouldn't involve any chain rule. If you mean something else, the notation is very weird.

I meant d/dx f(u). I messed up on the first, and then copied it for the others.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

Word. Does the explanation of why make sense?
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Re: Help with the Chain Rule

Yeah, I guess so. I bet it will make more sense when I solve more problems and see how it hows for equations that are structured differently. Still, I do not understand how is a corner/ cusp different than a curve, and what is a tangent line.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

A corner / cusp is a feature of a curve. You would say things like, "This curve has a cusp here," or, "This curve is continuous but not smooth, look, it has a 90 degree corner right there."

A tangent to a curve f at a point p is one which coincides with the curve at that point, and has the same slope as the curve / points in the same direction as the curve does at that point.

In the calculus sense, you can imagine the tangent as the limit of secant lines. If you took two points on a curve, f(p) and f(q), you could draw a line connecting them (the name of that line is the secant.) If you slide q towards p, and by slide, I mean take the limit as q approaches p, your secant lines approach what will be called the tangent line.

You took physics before this, right? If you use your physical intuition for instantaneous velocity as the mathematical benchmark for tangent, you are in the right place. (Though the magnitude of velocity means something that doesn't correspond generally. For tangents, we will either want to draw an infinite line or talk about a unit vector.)
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

jewish_scientist
Posts: 1047
Joined: Fri Feb 07, 2014 3:15 pm UTC

### Re: Help with the Chain Rule

I understand what a corner, cups, and tangent are; what I want to know is what their mathematical definitions are.
"You are not running off with Cow-Skull Man Dracula Skeletor!"
-Socrates

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

Corner, I don't know? I think it's where you just have a piecewise definition of the function.

Cusps are point in a curve which are continuous, and a derivative from the left and the right exist, but they don't agree. Like, if you looked at a { shape, the point at the tip would have two different tangents you could define, whether you are coming from the top or the bottom.

A tangent to a curve f at a point p is one which passes through f(p) and has slope equal to the derivative. You can use y = mx+ b with m = df/dx evaluated at p if you are in calc 1. If you want to worry about vertical lines, then you need to use the parametric form, and then the derivative will be well defined and you can draw your vertical line.
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?

cyanyoshi
Posts: 419
Joined: Thu Sep 23, 2010 3:30 am UTC

### Re: Help with the Chain Rule

I think doogly may be mixing up the terms a little bit. Or at least, it is different than what I was taught.

To understand the mathematical definitions, you need to be pretty familiar with the concept of limits, which can get surprisingly technical, but hopefully makes some intuitive sense. The derivative of the function f(x) at point a is the limit as x approaches a of the ratio (f(x)-f(a))/(x-a). If you think you understand what a tangent line is, the derivative is just the slope of the tangent line at any point along the curve. No big deal. In fact, this is basically the mathematical definition of a tangent line of the curve f(x) at point a: it's the line that passes through the point (a,f(a)) that has the same derivative as f(x) at point a.

Now that the TLDR of differential calculus is out of the way, you can define left and right derivatives by choosing values of x that are only to the left or right of the value a, respectively. A corner is where the left and right derivatives exist (or one is infinite), but they aren't equal to each other. For example, the absolute value function has a corner at x=0 because it has derivative +1 to the right of the origin, and -1 to the left. Since the curve looks like the line y=x for x>0 and y=-x for x<0, you can't really assign a single tangent line at x=0.

A continuous function has a cusp at point a if its left and right derivatives are both infinite, but they have different signs. Let's look at the curve y=|x|^(2/5). The curve becomes more and more vertical near the origin (the derivative approaches +/- infinity as x approaches 0). However, the derivative approaches +infinity from the right and -infinity from the left. Hence the cusp.

A definition I find more helpful comes from a bug crawling along the curve at constant speed. If the bug's direction always changes smoothly and gradually, then there is a tangent line at every point in the curve. The bug passes through a cusp if it must make a sudden 180° turn, and it passes through a corner if it has to turn by any other angle.

doogly
Dr. The Juggernaut of Touching Himself
Posts: 5538
Joined: Mon Oct 23, 2006 2:31 am UTC
Location: Lexington, MA
Contact:

### Re: Help with the Chain Rule

Word, those are the right words.
LE4dGOLEM: What's a Doug?
Noc: A larval Doogly. They grow the tail and stinger upon reaching adulthood.

Keep waggling your butt brows Brothers.
Or; Is that your eye butthairs?