Post

Notations-1

To keep consistency for the readers and my sanity, this article will be used as a reference to introduce notations for the following articles:

The common notations are related to random variables, probability density function (PDF) and probability mass function (PMF).

NotationSymbol
Random VariableX
Sample Value of a Random Variablex
Set of Possible Sample Values of XX
Probability Distribution (Both PDF and PMF)p(x)
Probability Mass Function (PMF)PX(x)
Probability Density Function (PDF)pX(x)
Expectation over Distribution pExp

The notation p(x) is used as a general placeholder for probability distributions. If a statement is true for both discrete random variables and continuous random variables to keep it general, I will use p(x) and I hope readers will be able to distinguish based on the context. For example, if I use a summation symbol with p(x), readers can assume that statement or equation can be generalized to an integration symbol for continuous random variables.

Other Notations

  • log : Assume natural logarithm or log base-e. Most resources you will come across for these topics, tend to use base- 2 logarithm, but in the machine learning domain, it is common practice to use a base- e logarithm.

  • I(X;Y): You can find more information about using a semicolon here. I might have as well used I(X,Y) instead.

  • KL Divergence: Even though it depends on the context of what I am talking about, in general, consider p as the true/target distribution which we are trying to approximate and q as the parameterized distribution. Most of these are only applicable when talking about variational inference.

This post is licensed under CC BY 4.0 by the author.