Velamuvi AIAcademy

Lesson 2 of 12 · Numbers and Algebra

Numbers and Operations

Refresh the basics: kinds of numbers, order of operations, negative numbers, fractions, percentages, absolute value, rounding and why computers get 0.1 + 0.2 slightly wrong.

  • 8 min read
  • Beginner
  • Reviewed
In this lesson you will learn
  • The kinds of numbers: whole numbers, integers and real numbers
  • The order of operations, so every calculation is read the same way
  • How to work with negative numbers, fractions and percentages
  • What absolute value is
  • Why computers sometimes give tiny rounding errors

Kinds of numbers

Maths sorts numbers into families. You will see these names in AI papers and documentation.

In Python, whole numbers are int and decimal numbers are float.

The order of operations

Is 2 + 3 × 4 equal to 20 or 14? Everyone must agree, so there is a fixed order.

StepOperationPython
1Brackets( )
2Orders, also called powers**
3Division and Multiplication, left to right/ *
4Addition and Subtraction, left to right+ -
Try it yourselfPython
print(2 + 3 * 4)          # multiplication first: 2 + 12
print((2 + 3) * 4)        # brackets first: 5 * 4
print(2 * 3 ** 2)         # power first: 2 * 9
print(10 - 4 - 3)         # left to right: 6 - 3

When in doubt, add brackets. They cost nothing and make your meaning clear.

Negative numbers

The sign rules for multiplying and dividing:

CalculationResultExample
positive × positivepositive3 × 2 = 6
negative × positivenegative-3 × 2 = -6
negative × negativepositive-3 × -2 = 6
Try it yourselfPython
print(-3 * 2)
print(-3 * -2)
print(5 - 8)
print(-2 ** 2)       # careful: this is -(2 ** 2)
print((-2) ** 2)

The last two lines show a classic trap: the power is done before the minus sign, so write brackets when you mean "minus two, squared".

Absolute value

Absolute value is how you measure how wrong a prediction is, without caring whether it was too high or too low:

Try it yourselfPython
actual = 100
for predicted in [90, 110, 100]:
    error = predicted - actual
    print(f"predicted {predicted}: error {error}, size of error {abs(error)}")

Fractions, decimals and percentages

These are three ways of writing the same idea: a part of a whole.

FractionDecimalPercentage
1/20.550%
1/40.2525%
3/40.7575%
1/100.110%
Try it yourselfPython
correct = 172
total = 200

accuracy = correct / total
print("As a decimal:", accuracy)
print("As a percentage:", accuracy * 100, "%")
print(f"Formatted: {accuracy:.1%}")      # :.1% shows a percentage with 1 decimal place
Try it yourselfPython
old_error = 0.40
new_error = 0.30
change = (new_error - old_error) / old_error * 100
print(f"The error changed by {change:.0f}%")

Rounding, and a computer surprise

Run this:

Try it yourselfPython
print(0.1 + 0.2)
print(0.1 + 0.2 == 0.3)
print(round(0.1 + 0.2, 2))

This is not a Python bug: every programming language does it. The fix is to round before showing results, and never to test decimals for exact equality. Python's math.isclose() checks whether two numbers are close enough.

Common mistakes

Practice exercise

Turn raw test results into accuracies and a percentage change, the way model reports do.

Your solutionPython
# A model's test results over three weeks
week1_correct, week1_total = 150, 200
week2_correct, week2_total = 168, 200

# 1. Work out each week's accuracy as a percentage.
# 2. Work out the percentage change from week 1 to week 2.
# Print each value rounded to 1 decimal place, exactly like this:

# Output must be exactly:
# Week 1 accuracy: 75.0%
# Week 2 accuracy: 84.0%
# Change: 12.0%
Knowledge check

What is 2 + 3 × 4?

Summary

  • Natural numbers count, integers include negatives, and real numbers fill the whole number line.
  • The order of operations: brackets, powers, multiply and divide, add and subtract.
  • Negative times negative is positive. Absolute value measures size without sign.
  • Fractions, decimals and percentages are three views of the same thing.
  • Computers have tiny floating-point errors, so round results and avoid == on decimals.

Next: variables and equations, the language for describing unknown values.

Finished reading? Tick it off to track your progress.