Module 4 · Lesson 4.2

Differentiation Rules — Power, Product & Quotient

Use the power, product, and quotient rules as ready-made shortcuts that replace writing out the limit definition every time.

Intuition

Computing a derivative from the limit definition works, but it quickly becomes tedious. The same algebraic patterns keep appearing: powers drop an exponent, products split into two terms, quotients follow a fixed four-piece recipe. Those recurring patterns are turned into named rules so you never have to re-derive them.

In machine learning the loss function is almost always a complicated combination of powers, products, and quotients. Automatic differentiation engines apply exactly these rules under the hood to obtain the gradient that updates the network weights.

Core concept

Differentiation rules are algebraic shortcuts. Once you recognize the form of an expression—pure power, product of two functions, or quotient of two functions—you apply the matching formula and obtain the derivative without returning to the limit definition.

$f(x)=x^{3}$$f'=3x^{2}$
At $x=2$$f'(2)=12$
$g=x^{2}\cdot x$same as $x^{3}$
Product rule$2x\cdot x+x^{2}\cdot 1=3x^{2}$
Checkmatches power rule

Interactive Lab

Pick a rule, adjust the parameters, and watch the symbolic derivative and its numeric value at a chosen point update instantly.

$f(x)=x^{3}$
$f'(x)=3x^{2}$
At $x=2$: $f'=12$

Change the rule or the parameters. The box shows both the symbolic derivative produced by the rule and its numeric value at the chosen $x$. These are the same shortcuts an autodiff engine applies when differentiating a loss function.

Key Idea

$$ \begin{aligned} (x^{n})' &= n\,x^{n-1} \\[6pt] (uv)' &= u'v + uv' \\[6pt] \Bigl(\frac{u}{v}\Bigr)' &= \frac{u'v - uv'}{v^{2}} \end{aligned} $$

The power rule drops the exponent and multiplies by the old power. The product rule splits a product into two terms, each differentiating one factor while leaving the other alone. The quotient rule follows the analogous pattern for a fraction, with a minus sign and a squared denominator. Together they let you differentiate any rational combination of powers without returning to limits.

$n$The exponent in a pure power $x^{n}$
$u,v$Differentiable functions being multiplied or divided
$u',v'$The derivatives of those functions
$v^{2}$Squared denominator that appears in the quotient rule

Think Further

  1. Why does the product rule contain two terms instead of simply $u'v'$?
  2. When an AI framework computes the gradient of a loss that contains a soft-max, which of these three rules is applied most often?
  3. Show that the quotient rule can be derived from the product rule by writing $\frac{u}{v}=u\cdot v^{-1}$ and using the power rule on $v^{-1}$.
Show suggested answers
  1. Both factors change. One term measures the change in $u$ while $v$ is held, and the other measures the change in $v$ while $u$ is held.
  2. The chain rule is used most often because softmax is nested inside the larger loss function; product and quotient rules may appear inside that chain.
  3. Differentiate $uv^{-1}$: $u'v^{-1}+u(-v^{-2}v')=\frac{u'v-uv'}{v^2}$.