Course contentsShow
← Machine Learning and Deep Learning
Lesson 59 of 3,5381. Mathematical Foundations for Machine LearningFree lesson

Probability Mass Functions

Define and work with PMFs for discrete random variables to specify probability distributions.

Probability Mass Functions

What you'll learn: How to use Probability Mass Functions (PMFs) to describe the complete probability distribution of a discrete random variable.

What Is a PMF?

A Probability Mass Function assigns a probability to each possible value that a discrete random variable can take. Think of it as a lookup table: for each outcome, the PMF tells you exactly how likely it is to occur.

Remember from the previous lesson that discrete random variables take on specific, countable values (like rolling a die: 1, 2, 3, 4, 5, or 6). The PMF, usually written as P(X = x) or simply p(x), maps each value to its probability.

Key Properties

A valid PMF must satisfy two rules:

  1. Non-negativity: All probabilities are between 0 and 1: 0 ≤ p(x) ≤ 1
  2. Normalization: The sum of all probabilities equals 1: Σ p(x) = 1

Example: Coin Flips

Let's say X represents the number of heads in two coin flips. Possible values are {0, 1, 2}.

  • P(X = 0) = 1/4 (both tails)
  • P(X = 1) = 2/4 (one head, one tail)
  • P(X = 2) = 1/4 (both heads)

Notice: 1/4 + 2/4 + 1/4 = 1 ✓

Why PMFs Matter in ML

PMFs are foundational for:

  • Classification problems: Each class has a probability
  • Understanding data distributions: Knowing how frequent each value is
  • Building models: Many algorithms assume or learn probability distributions

The PMF gives you the complete picture of a discrete random variable's behavior—every possible outcome and its likelihood.

Key Takeaway: A Probability Mass Function specifies the exact probability for each value a discrete random variable can take, and all these probabilities must sum to 1.