Tensor by Tensor

MODULE 01 · LESSON 02

Derivatives

A derivative tells how fast an output changes near one input value.

PLAIN-LANGUAGE INTRODUCTION

What is this?

A derivative tells how fast an output changes near one input value.

One simple example

Use h(t) = t² at t = 3. Its derivative there is 6.

What goes in?

A function and one point. Here, use h(t) = t² and t = 3.

What comes out?

A local slope. Here, the slope is 6.

Why does it matter?

Training needs local slopes to choose small parameter changes.

What is it not?

A derivative is not the function value. Here, the value is 9.

WORK THROUGH THE IDEA

See the idea in more detail

  1. A slope compares an output change with an input change. A derivative is the slope at one point.
  2. For h(t) = t², the derivative is h′(t) = 2t. The mark means derivative.
  3. At t = 3, compute 2 × 3 = 6. This is the local slope.
  4. A move from 3 to 3.1 predicts a rise near 6 × 0.1 = 0.6. The actual rise is 9.61 − 9 = 0.61.
  5. Common mistake: the slope 6 is not the height. The height at t = 3 is 9.
Open the detailed notes ↗