\documentclass{article}
\usepackage[margin=1in]{geometry}
\usepackage{hyperref}
\usepackage{amsmath,amsfonts,amssymb,amsthm,commath,dsfont}
\usepackage{enumitem}
\usepackage[round]{natbib}
\usepackage[dvipsnames]{xcolor}
\newcommand{\red}[1]{{\color{red} #1}}
\title{ML Theory --- Project Proposal}
\author{djhsu + mjt \red{[list up to two members here]}}
\date{\textbf{Due:} 3pm on Wednesday, November 14. \red{[~delete this line\dots~]}}
\begin{document}
\maketitle
\begin{description}
\item[Interests.]
\red{[List 1-5 sentences in this section. For example:
\begin{quote}
We have technical background in linear algebra, spectral methods,
latent variable models, tensors, convex analysis, functional analysis,
probability, and optimization.
We are interested in sampling, optimization, representation, and generalization
in the non-convex setting.
\end{quote}
]}
\item[Proposal.]
\red{[ This section is also 1-5 sentences, but it needs to pass a ``sniff test'';
if you write a random sentence, don't expect credit. Here are some project suggestions:
\begin{itemize}
\item
Provide a \emph{cohesive} survey of some collection of papers. Remember that
your presentation is 2 slides and your handin is 1-2 pages (2 pages is a \emph{maximum}).
Therefore you can't just summarize, you need to purify and unify.
\item
Provide a major simplification or reimagining of a single paper.
For instance, I simplified the classical ``universal approximation''
proofs based on Stone-Weierstrass by using the activation
$\exp(\cdot)$, amongst other things \citep{nn_stone_weierstrass}.
I learned a lot from this!
\item
Study one (or a few) fundamental classical papers,
``modernize'' their presentation and proofs,
and tie them to present literature.
\item
Solve an open problem.
\end{itemize}
Your project needs to include \emph{some} theoretical content. ]}
\end{description}
\bibliographystyle{plainnat}
\bibliography{proposal}
\noindent\red{[Include 3-10 references via proper bibtex.]}
\end{document}