Simplifying and automating LaTeX with `shortex.sty`

Using LaTeX has a number of big benefits when writing papers:

it separates content (your words) and format (how they are laid out on a page)
it is stored in plaintext, which is useful for version control
it is extendable, a huge base of helpful packages exists
it keeps track of references, bibliography citations, links
it is currently the only decent option for typesetting math
it makes professional-looking papers by default
it is basically required by every conference/journal in ML/stats

But it also has a number of drawbacks. The shortex.sty style file is designed to help make typesetting a little less painful and LaTeX code a little easier to read. It provides:

automatic reference typing using cleveref
automatic equation numbering using autonum
a few nice commands for adding comments inline or in the margins
shortened commands for align environments
short commands for commonly used math symbols
a collection of oft-imported packages

Grab a copy of the style file from the shortex repository, and use the guide below to learn how to use the package. Note for those who don't read through the rest of the post: the automatic equation numbering in shortex requires you to compile your document 4 times.

By the way: if, in your travels, you find a really neat package that makes LaTeX less annoying to use, let me know and I'll consider adding it to shortex.sty!

Note: the shortex package is currently undergoing a big improvement, with the help of the amazing Jonathan Huggins and Jeffrey Negrea. Stay tuned for updates to this blog post.

Automatic reference typing

Typically to use a reference in LaTeX you have to write the name of the type of reference yourself. For example, if you want to reference a figure, you would use something like

In Figure \ref{fig:the_figure}, you can see...

or for multiple figures, you might use

Figures \ref{fig:the_figure}, \ref{fig:the_other_figure}, and \ref{fig:the_third_one} show that...

The cleveref package simplifies this process significantly. Use the \cref command to automatically typeset the names of the objects you're referencing (including properly handling multiple references). The above two examples become

In \cref{fig:the_figure}, you can see...

and

\cref{fig:the_figure,fig:the_other_figure,fig:the_third_one} show that...

This works for many different reference types (Figure, Algorithm, Equation, Table, etc), and can be extended if needed. See the cleveref package for more.

Simplified equations and automatic equation numbering

Typically when you typeset equations, you have to choose between $...$ , $$...$$, \begin{align}...\end{align}, \begin{aligned}...\end{aligned}, \begin{equation}...\end{equation}, not to mention starred versions of those environments and \nonumber/\notag commands, depending on whether/where you want equation numbers, display or in-text math, etc. This leads to verbose, inconsistent code.

The shortex package provides two major improvements. First, we replace the align environment with a much less verbose \[ ... \] syntax,

\[
   A &= B + C\\
   D &\leq E
\]

And second, there are only two commands you need to use: single dollar signs for in-text math,

This is some in-text math $a+b=c$

and square brackets for display math mode,

\[
   a+b = c \label{eq:the_equation}
\]

The autonum package automatically decides which equations to provide numbers based on which equations you reference. For example, in the above display math, I used the label eq:the_equation. If I use the command \cref{eq:the_equation} somewhere in the document, that equation will automatically be assigned a number. If not, it won't get a number. See the autonum package for more.

Note: You must compile your document 4 times for this to work properly.

Convenient math commands

The shortex.sty file has a huge list of shortened commands for mathematical symbols. I'll list a few of the most common ones I use here, but please go through the style file to see the full list.

You can see the difference in clarity below. Here is some basic LaTeX math syntax:

\begin{align}
X_n &\in \mathcal{X}\\
X_n &\overset{\text{p}}{\to} X\\
\mathbb{E}\left[X\right] &= \int_0^1 6\frac{\mathrm{d} 4y^2}{\mathrm{d} y}\mathrm{d}y\\
\mathbb{P}\left(X_n \in \mathcal{A}\right) \leq 
X_n \sim \mathcal{N}(\hat{\mu}, \hat{\sigma}).
\end{align}

And here is the exact same expression using shortex.sty syntax:

\[
X_n &\in \mcX\\
X_n &\convp X\\
\EE\left[X\right] &= \int_0^1 6\der{4y^2}{y}\dee y\\
X_n \dist \distNorm(\hmu, \hSigma)
\]

Sets

\reals      % real numbers
\rats       % rational numbers
\nats       % natural numbers
\comps      % complex numbers

General Purpose

\defined    % := for "defined as"
\defines    % =: for "defines"
\argmax     % argmax
\argmin     % argmin
\st         % s.t., such that

Theorem Environments

\bthm ... \ethm  % unnumbered theorem
\blem ... \elem  % unnumbered lemma
\bprop ... \eprop  % unnumbered proposition
\bcor ... \ecor  % unnumbered corollary

and to get the numbered version of these, just insert an n :

\bnthm ... \enthm  % unnumbered theorem
\bnlem ... \enlem  % unnumbered lemma
\bnprop ... \enprop  % unnumbered proposition
\bncor ... \encor  % unnumbered corollary

You can create a proof block below a result statement using the \bprf ... \eprf block:

\bthm
here is the theorem statement.
\ethm
\bprf
here is the proof
\eprf

You can also put a proof later on in the document (e.g. in the appendix) by using the \bprfof{...} ... \eprfof block and referencing a labelled and numbered theorem

\bnthm \label{thm:the_theorem}
here is the theorem statement.
\enthm

% (more of your document here)

\bprfof{\cref{thm:the_theorem}}
here is the proof
\eprf

Probability / Stats

Text-mode shortcuts:

\iid        % i.i.d.
\as         % a.s.
\aev        % a.e.v.

Math-mode shortcuts:

\convas     % converges almost surely to
\convp      % converges in probability to
\convd      % converges in distribution to
\eqd        % equal in distribution to
\eqas       % equal almost surely to

\dist       % distributed as
\distind    % distributed independently as
\distiid    % distributed iid as

\Pr         % the probability of
\EE         % expectation
\var        % variance
\cov        % covariance

Distributions (there are quite a few more in shortex.sty, and you can define new ones using the \distNamed command)

\distBern
\distNorm
\distUnif
\distGam
\distPoiss
\distBeta
\distExp
\distMulti

Calculus

\dee              % a math roman numeral "d", useful for integrals, e.g. "dt"
\der{...}{...}    % derivatives, e.g. \der{F}{t} will show dF/dt
\dder{...}{...}   % second derivatives
\D{...}{...}      % partial derivatives
\DD{...}{...}     % partial second derivatives

Accented symbols

To insert a bar over most characters, just put a \b in front of it. For example,

\ba       % overbar a
\bG       % overbar G
\balpha   % overbar alpha

This exact same pattern applies to many different style changes. Here is a list of useful ones:

\h...     % hat
\t...     % tilde
\mc...    % mathcal
\bh...    % bold hatted
\wh...    % wide hatted
\bt...    % bold tilde
\bi...    % bold italic
\mf...    % mathfrak

Each of these accents does not work for all characters, but it works for a lot of them (even greek letters).

Lists

To typeset an unnumbered list, use \bitems...\eitems:

\bitems
\item Here is an item
\item Here is another
\eitems

To typeset a numbered list, use \benum...\eenum:

\benum
\item Here is an item
\item Here is another
\eenum

Simplifying and automating LaTeX with shortex.sty