Lesson 11 of 12 · Reading Maths
Sets and Intervals
Learn set notation (∈, ⊆, ∪, ∩, ∅), the number sets ℕ, ℤ and ℝ, and interval notation with square and round brackets, with Python sets and range checks for every idea.
- 7 min read
- Beginner
- Reviewed
- What a set is and how to read set notation
- Union, intersection and subsets
- The number sets ℕ, ℤ and ℝ
- How to read intervals such as [0, 1] and (0, 1)
- How each idea looks in Python
Groups of things
Python has sets built in, and they behave exactly like maths sets:
A = {1, 2, 3, 3, 2}
print(A) # repeats disappear
print(2 in A) # 2 ∈ A
print(7 not in A) # 7 ∉ A
print(len(A)) # how many elementsCombining sets
python_students = {"Ada", "Ravi", "Mei", "Sara"}
sql_students = {"Ravi", "Sara", "Tom"}
print("Union: ", sorted(python_students | sql_students))
print("Intersection:", sorted(python_students & sql_students))
print("Difference: ", sorted(python_students - sql_students))
print("Empty set: ", set())sorted() is used only to print the names in a fixed order, because sets have no order of their own.
Subsets
labels_in_test = {"cat", "dog"}
all_labels = {"cat", "dog", "bird"}
print(labels_in_test <= all_labels) # labels_in_test ⊆ all_labels
print(all_labels <= labels_in_test)This is a real check in machine learning: every label in your test data should be one the model was trained on.
The number sets
Some sets are so important they have their own double-struck letters:
You will often see these in AI papers in phrases like x ∈ ℝ, meaning "x is a real number", or x ∈ ℝⁿ, meaning "x is a list of n real numbers". So "an input x ∈ ℝ⁷⁸⁴" means the input is a list of 784 real numbers, such as the pixels of a 28 × 28 image.
Intervals
| Notation | Meaning | In Python |
|---|---|---|
[0, 1] | from 0 to 1, including both ends | 0 <= x <= 1 |
(0, 1) | between 0 and 1, not including either end | 0 < x < 1 |
[0, 1) | including 0, but not 1 | 0 <= x < 1 |
(0, ∞) | every number above 0 (∞ means infinity) | x > 0 |
def in_closed(x, low, high): # [low, high]
return low <= x <= high
def in_open(x, low, high): # (low, high)
return low < x < high
for x in [0, 0.5, 1]:
print(f"x = {x}: in [0, 1]? {in_closed(x, 0, 1)} in (0, 1)? {in_open(x, 0, 1)}")Intervals tell you the allowed values of things in AI: a probability is in [0, 1], the output of sigmoid is in (0, 1) because it never quite reaches 0 or 1, and a learning rate is in (0, ∞).
Common mistakes
Use set operations to catch a data leak, and interval checks to catch impossible probabilities.
train_ids = {101, 102, 103, 104, 105}
test_ids = {104, 105, 106, 107}
# 1. Print the ids that appear in BOTH sets, sorted (these are a data leak!).
# 2. Print how many distinct ids there are in total.
# 3. For each probability below, print whether it is in the interval [0, 1].
probabilities = [0.3, 1.2, 0, -0.1]
# Output must be exactly:
# Leaked ids: [104, 105]
# Total distinct ids: 7
# 0.3 True
# 1.2 False
# 0 True
# -0.1 False
Which numbers belong to the interval (2, 5]?
Summary
- A set holds distinct elements.
x ∈ Ameans x is in A. - Union ∪ is either, intersection ∩ is both, and ∅ is the empty set. In Python:
|,&andset(). - A ⊆ B means every element of A is also in B.
- ℕ, ℤ and ℝ are the natural numbers, integers and real numbers.
x ∈ ℝⁿmeans a list of n real numbers. - In an interval, square brackets include the end point and round brackets leave it out.
Next: putting it all into practice with maths in Python.
Finished reading? Tick it off to track your progress.