What is most preferable/acceptable notation on parameterized probability distribution and expectation?

122 Views Asked by At

I am coming from machine learning community, during reading papers, different authors use different notations for parameterized probability distribution and expectation. Thus, I am quite curious to see what is most preferable or acceptable notation to recommend from mathematics community.

There are a few examples as following

  1. A probability density function over $x$

    (a): $p(x)$

    (b): $P(x)$

    (c): $\mathbb{P}(x)$

    (d): $\mathrm{P}(x)$

For following questions, by default using $P(x)$.

  1. Given a probability distribution $P(x)$ parameterized by a vector $\theta$.

    (a): $P(x|\theta)$, I don't like this, because it confuses with conditional probability distribution.

    (b): $P(x; \theta)$, also not a big fun of it,

    (c): $P_{\theta}(x)$, personal favorite

For following question, by default, using $P_\theta(x)$

  1. Given a random variable $x$ with PDF $P(x)$ parameterized by $\theta$, the expectation operation

    (a): $E[x|\theta]$ or $\mathbb{E}[x|\theta]$

    (b): $E_\theta[x]$ or $\mathbb{E}_\theta[x]$

    (c): $E_{x\sim P_\theta(x)}[x]$ or $\mathbb{E}_{x\sim P_\theta(x)}[x]$

    (d): $E_{x\sim P_\theta(\cdot)}[x]$ or $\mathbb{E}_{x\sim P_\theta(\cdot)}[x]$