Chapter 1: Tiny math map
The handful of symbols this whole book is written in.
None of this is advanced. It is a map for reading the examples, and you can flip back to it whenever a line of symbols looks strange.
| What you see | What it means |
|---|---|
0 | no, off, false, absent |
1 | yes, on, true, present |
[1, 0] | two inputs: the first one is on, the second one is off |
(1, 0) | the same two numbers, drawn as a dot on graph paper |
> | "is bigger than?" — the question that turns a number into a yes/no answer |
≈ | "is about": 1 ÷ 3 ≈ 0.33 |
Σ | "add them all up" (a Greek capital S, for sum) |
One example, three costumes
This book keeps coming back to a single running example called the screen-time rule. The house rule is: you get screen time if you practiced piano OR finished your homework. In chapter 4 you'll teach a neuron that rule. For now, notice that one example from it can be written three ways.
As a table row:
piano = 1, homework = 0, answer = YES
As a line of code, with an input and a label:
x = [1, 0] y = 1
And as a graph dot: the point (1, 0) with a YES label next to it. Same information, three costumes.
Two words you'll see on every page
A weight is "how much this input matters." A bias is "how easy it is for the neuron to say yes before it sees any input." The neuron adds everything up, then asks one yes/no question: is the total big enough?
weight1 × input1 + weight2 × input2 + bias > 0 means "input 1 gets a vote, input 2 gets a vote, the bias adds a built-in lean, and the neuron asks whether the total crosses the yes line."
A table row, a dot on graph paper and a line of code are three ways to write the same example. When symbols pile up, translate them back into words.
Write the input [0, 1] in words for the screen-time example. Which thing is off? Which thing is on?
In the screen-time example, the inputs are written [piano, homework]. Which dot on graph paper is "homework done, piano not done"?
Order matters. Piano comes first and it is off (0); homework comes second and it is on (1). So the dot is (0, 1).
A neuron has a big weight on input 1 and a weight of 0 on input 2. What does that tell you?
A weight is not the input itself. It is how much that input counts. Anything multiplied by a weight of 0 adds nothing to the total.
What is the job of the > 0 at the end of weight1 × input1 + weight2 × input2 + bias > 0?
Everything before the > is just arithmetic that produces one number. The comparison is what turns that number into a decision: above the line is YES, below is NO.