Bernoulli, Gaussian, exponential and beta side by side, each shaped by its own parameters.
From the chapter: Chapter 4: Probability
Glossary: bernoulli distribution, gaussian distribution, exponential distribution, beta distribution
Transcript
A handful of named distributions appear again and again in machine learning. Each is shaped by a small set of parameters that control its behaviour.
The Bernoulli distribution is the simplest. It assigns probability p to one outcome and one minus p to the other. Watch the bars rebalance as p slides from zero to one.
The Gaussian or normal distribution is the bell curve. It is governed by two numbers: a mean that shifts the peak, and a standard deviation that controls its width. Most of the probability mass sits within two standard deviations of the mean.
The exponential distribution describes the time between events in a memoryless process. A single rate parameter controls how quickly it decays.
The beta distribution lives on the unit interval and is the workhorse of probabilities of probabilities. Its two parameters can make it U-shaped, bell-shaped, skewed left, or skewed right.
These four shapes, with their handful of parameters, model a great amount of the world.