Mean, variance and spread

Why it matters

The expected value is the long-run average: expected revenue per visitor, expected loss over a training set, expected reward in reinforcement learning. Variance measures how much individual results scatter around it, which is what separates a steady channel from a lottery ticket, and a stable training run from a noisy one. Training a model literally means minimizing an expected loss.

What you'll be able to do

• Compute an expected value as a probability-weighted average
• Measure spread with variance and standard deviation
• Use the rules for scaling and adding random variables
• Compare options with the same average but different risk

Before this lesson

• Distributions

Next in this phase: The normal distribution