path: root/reproducible-paper.tex

\documentclass[9pt]{beamer}

%% Beamer settings.
%\setbeamertemplate{footline}[frame number]

%% Packages to import.
\usepackage{tcolorbox}          %For a color-box.
\usepackage{textcomp}           %For a copyright sign.

%% To simplify arXiv links
\newcommand{\arxivlink}[1]{{\footnotesize
    (\textcolor{blue}{\href{https://arxiv.org/abs/#1}{arXiv:#1}})}}

%% Set the title
\title{\LARGE \textbf{BIG} data, \textbf{BIG} responsibility:\\
  \small Template/framework for reproducible scientific projects/papers}

%% Set the author
\author{Mohammad Akhlaghi\\\vspace{2mm}\footnotesize Instituto de
  Astrof\'isica de Canarias ({\scriptsize IAC}),\\Tenerife, Spain\\
  \vspace{0.5cm} \raisebox{+0.8\height}{\includegraphics[width=2cm]{img/ull.png}}{ }{
  }\includegraphics[width=1.5cm]{img/iac.png}{ }{ }{
  }\includegraphics[width=2cm]{img/sundial.png}\vspace{0.5cm}
}

%% Set the date and insitutional logos.
\date{\scriptsize \href{https://www.astro.rug.nl/~sundial/MidtermMeeting.html}{SUNDIAL Midterm meeting}, June 5th, 2019\\ Ghent, Belgium}

%% For a wider writing width.
\newcommand\Wider[2][3em]{%
\makebox[\linewidth][c]{%
  \begin{minipage}{\dimexpr\textwidth+#1\relax}
  \raggedright#2
  \end{minipage}%
  }%
}

%% TiKZ
\usepackage{tikz}
\usetikzlibrary{graphs}
\usetikzlibrary{positioning}
\tikzset{ bbox/.style={
  rectangle,
  minimum width=2.5cm,
  rounded corners=2mm,
  very thick,draw=black!50,
  top color=white,
  bottom color=black!20 } }

\tikzset{ rbox/.style={
    rectangle,
    dotted,
    minimum width=2.5cm,
    rounded corners=2mm,
    very thick,draw=red!50!black!50,
    top color=white,
    bottom color=red!50!black!20 } }

\tikzset{ gbox/.style={
    rectangle,
    minimum width=2.5cm,
    very thick,
    draw=green!50!black!50,
    top color=white,
    bottom color=green!50!black!20 } }




















\begin{document}

  \begin{frame}
    \titlepage
  \end{frame}





  \begin{frame}{Necessity of (exactly) reproducible research}
    \begin{itemize}
      \setlength\itemsep{0.7cm}
    \item To be considered \alert{scientific}, any result has to be
      reproducible.
    \item The tsunami of data, fast internet, and high processing
      power have made it very easy to \alert{promptly arrive at a
        result}.
    \item But these factors have also greatly increased the
      \alert{complexity} of an analysis. Making it impossible to
      exactly describe all steps in a traditional published paper.
    \item Most scientific papers thus \alert{ignore some ``details''}
      (as they interpret it).
    \item But due to the complexity, even a small deviation from the
      exact result, can be due to many different parts of the
      analysis. Hence, its \alert{critical to exactly reproduce} a
      result.
    \end{itemize}
  \end{frame}




  \newcommand{\nodeopacity}{1}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\paperinit}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\sver}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\srep}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\dver}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\ddver}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\confopt}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\confenv}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\db}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\calib}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\corr}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\runord}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\runopt}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\humanerr}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\depupdate}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\coauth}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\varsinpaper}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\recordinfo}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\softcite}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\prevchange}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}
  \newcommand{\paperfinal}{}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}

  %% Don't show the happy scientist any more.
  \let\paperfinal\undefined
  \let\paperinit\undefined





  \begin{frame}{Science is a tricky business}
    \includegraphics[width=\linewidth]{img/nature-cartoon.jpg}

    \vspace{-0.2cm}

    {\tiny Image from nature.com
      (``\href{https://www.nature.com/articles/d41586-017-07522-z}{Five
        ways to fix statistics}'', Nov 2017)}

    \vspace{0.2cm}
    \begin{tcolorbox}
      \small Data analysis [...] is a human behaviour. Researchers
      who hunt hard enough will turn up a result that fits
      statistical criteria, but their \alert{discovery} will
      probably be a \alert{false positive}.

      \hfill Five ways to fix statistics, Nature, 551, Nov 2017.
    \end{tcolorbox}
  \end{frame}





  \begin{frame}{Necessity of (exactly) reproducible research}
    \begin{tcolorbox}[title=Don't forget that:]
      \centering Science is defined by its METHOD, \alert{not} its
      result.
    \end{tcolorbox}

    \vspace{0.5cm}
    \begin{itemize}
      \setlength\itemsep{0.6cm}
    \item The software(s) used, configuration file(s), the order of
      steps taken, along with the input data are necessary for
      reproducibility.
    \item \alert{A solution} is proposed here, which if adopted from
      the start, can greatly \alert{simplify a scientific research
        project} and \alert{allow full/exact reproducibility} once it
      is published.
    \item In the next slides, we'll review the template from the
      highest level (final research paper) to the lowest (setting up
      the research environment).
    \end{itemize}
  \end{frame}



  \renewcommand{\nodeopacity}{0.3}
  \begin{frame}{General outline of a project} \include{tex/project-graph} \end{frame}





  \begin{frame}{Values in final report/paper}
    All necessary analysis/processing \alert{input} and \alert{output}
    values are written into the final report as \LaTeX{} macros. Shown
    here is a portion of the \textsf{NoiseChisel} paper and its source
    (\textcolor{blue}{\small\href{https://arxiv.org/abs/1505.01664}{arXiv:1505.01664}}).

    \vspace{1.2cm}
    \includegraphics[width=\linewidth]{img/reproducible-latex.png}
  \end{frame}

  \begin{frame}{Values in final report/paper}
    All necessary analysis/processing \alert{input} and \alert{output}
    values are written into the final report as \LaTeX{} macros. Shown
    here is a portion of the \textsf{NoiseChisel} paper and its source
    (\textcolor{blue}{\small\href{https://arxiv.org/abs/1505.01664}{arXiv:1505.01664}}).

    \vspace{1.2cm}
    \includegraphics[width=\linewidth]{img/reproducible-latex-highlighted.png}
  \end{frame}


  \begin{frame}{Values come from a single file}
    All the \LaTeX{} macros (processing inputs and outputs) come from
    a \alert{single file}. This file is the \alert{final product} of
    the analysis steps.

    \begin{center}
      \includegraphics[width=0.8\linewidth]{img/reproducible-macros.png}
    \end{center}
  \end{frame}



  \begin{frame}{Values come from a single file}
    All the \LaTeX{} macros (processing inputs and outputs) come from
    a \alert{single file}. This file is the \alert{final product} of
    the analysis steps.

    \begin{center}
      \includegraphics[width=0.8\linewidth]{img/reproducible-macros-highlighted.png}
    \end{center}
  \end{frame}


  \begin{frame}{Values written during analysis}
    Various steps of the analysis write the macro values as soon as
    they are calculated internally.

    \begin{center}
      \includegraphics[width=0.8\linewidth]{img/reproducible-write-macro.png}
    \end{center}
  \end{frame}


  \begin{frame}{Values written during analysis}
    Various steps of the analysis write the macro values as soon as
    they are calculated internally.

    \begin{center}
      \includegraphics[width=0.8\linewidth]{img/reproducible-write-macro-highlight.png}
    \end{center}
  \end{frame}


  \begin{frame}{Reproducible science: Template is managed through a Makefile}
    \small
    \begin{columns}
      \column{5.5cm}

      All steps (downloading and analysis) is managed by Makefiles
      (example from
      \textcolor{blue}{\small\href{https://doi.org/10.5281/zenodo.1164774}{zenodo.1164774}}):

      \begin{itemize}
        \setlength\itemsep{0.2cm}
      \item Unlike a script which always starts from the top, a
        Makefile \alert{starts from the end} and steps that don't
        change will be left untouched (not remade).
      \item A single \emph{rule} can \alert{manage any number of
        files}. See the examples here where \textsf{NoiseChisel} and
        \textsf{MakeCatalog} are run separately on \alert{$\sim20$
          files} (different filters/fields) with a single rule.
      \item Make can identify independent steps internally and do them
        in \alert{parallel}.
      \item Make was \alert{designed for complex problems} with
        thousands of files (all major Unix-like components), so it is
        highly evolved and efficient.
      \item Make is a very \alert{simple} and \alert{small} language,
        thus easy to learn with great and free documentation (for
        example
        \textcolor{blue}{\href{https://www.gnu.org/software/make/manual/}{GNU
            Make's manual}}, usable to learn all implementations).
      \end{itemize}

      \column{5.5cm}
      \includegraphics[width=\linewidth]{img/reproducible-makefile.png}
    \end{columns}
  \end{frame}





  \begin{frame}{Predefined/exact software tools}
    \small
    \begin{columns}
      \column{5.5cm}
      \begin{tcolorbox}[width=\linewidth, boxsep=1pt, left=1pt, right=1pt,
                        top=1pt, bottom=1pt, title=Reproducibility \&
                        software]
        \footnotesize Reproducing the environment (specific
        \alert{software versions}, \alert{build instructions} and
        \alert{dependencies}) is also critically important for
        reproducibility.
      \end{tcolorbox}

      \begin{itemize}
        \setlength\itemsep{0.4cm}
      \item \emph{Containers} or \emph{Virtual Machines} are a
        \alert{binary black box}: just contain the environment, not
        how to set it up, or its history. They are also an overhead.

      \item This template \alert{installs fixed versions} of all
        necessary research software and their dependencies, down to
        the command-line shell, C compiler, POSIX tools and Python
        interpreter. It just avoids very low-level OS elements like
        the kernel or linker.

      \item Installs similar environment on \alert{GNU/Linux}, or
        \alert{macOS} systems.

      \item Works very much like a package manager (e.g.,
        \alert{\texttt{apt}} or \alert{\texttt{brew}}).
      \end{itemize}

      \column{5.5cm}
      \includegraphics[width=\linewidth]{img/software.png}
    \end{columns}
  \end{frame}





  \begin{frame}{Predefined/exact software tools}
    \small
    \begin{columns}
      \column{5.5cm}
      \begin{tcolorbox}[width=\linewidth, boxsep=1pt, left=1pt, right=1pt,
                        top=1pt, bottom=1pt, title=Reproducibility \&
                        software]
        \footnotesize Reproducing the environment (specific
        \alert{software versions}, \alert{build instructions} and
        \alert{dependencies}) is also critically important for
        reproducibility.
      \end{tcolorbox}

      \begin{itemize}
        \setlength\itemsep{0.4cm}
      \item \emph{Containers} or \emph{Virtual Machines} are a
        \alert{binary black box}: just contain the environment, not
        how to set it up, or its history. They are also an overhead.

      \item This template \alert{installs fixed versions} of all
        necessary research software and their dependencies, down to
        the command-line shell, C compiler, POSIX tools and Python
        interpreter. It just avoids very low-level OS elements like
        the kernel or linker.

      \item Installs similar environment on \alert{GNU/Linux}, or
        \alert{macOS} systems.

      \item Works very much like a package manager (e.g.,
        \alert{\texttt{apt}} or \alert{\texttt{brew}}).
      \end{itemize}

      \column{5.5cm}
      \includegraphics[width=\linewidth]{img/software-highlighted.png}
    \end{columns}
  \end{frame}


  \newcommand{\redbdir}{\textcolor{green!80!black}{/TEMPLATE/BUILD/DIRECTORY/software/installed/lib}}
  \begin{frame}{Dependencies are cleanly managed}

    \begin{itemize}
    \item All the software are configured and built to use the
      \alert{template's own builds}: indepenent of host system
      (\textcolor{green!80!black}{in green}).
    \item Template even builds a fixed GNU C Compiler (\alert{GCC}).
    \item Only extremely low-level dependencies (for example C library
      and Kernel) not built.
      \begin{itemize}
      \item GNU C library will also be added later (\alert{in red}).
      \end{itemize}
    \end{itemize}

    \vspace{0.5cm}
    \tiny\texttt{
      \$ ldd .local/bin/astnoisechisel\\
      \hspace{0.5cm}libgit2.so.26 => \redbdir/libgit2.so.26 (0x00007febb5232000)\\
      \hspace{0.5cm}libtiff.so.5 => \redbdir/libtiff.so.5 (0x00007febb51b8000)\\
      \hspace{0.5cm}liblzma.so.5 => \redbdir/liblzma.so.5 (0x00007febb5190000)\\
      \hspace{0.5cm}libjpeg.so.9 => \redbdir/libjpeg.so.9 (0x00007febb5153000)\\
      \hspace{0.5cm}z.so.1 => \redbdir/libz.so.1 (0x00007febb5136000)\\
      \hspace{0.5cm}wcs.so.6 => \redbdir/libwcs.so.6 (0x00007febb4fcc000)\\
      \hspace{0.5cm}cfitsio.so.8 => \redbdir/libcfitsio.so.8 (0x00007febb4caf000)\\
      \hspace{0.5cm}curl.so.4 => \redbdir/libcurl.so.4 (0x00007febb4c35000)\\
      \hspace{0.5cm}ssl.so.1.1 => \redbdir/libssl.so.1.1 (0x00007febb4b9b000)\\
      \hspace{0.5cm}crypto.so.1.1 => \redbdir/libcrypto.so.1.1 (0x00007febb48b5000)\\
      \hspace{0.5cm}gsl.so.23 => \redbdir/libgsl.so.23 (0x00007febb4626000)\\
      \hspace{0.5cm}gslcblas.so.0 => \redbdir/libgslcblas.so.0 (0x00007febb45e2000)\\
      \hspace{0.5cm}gnuastro.so.8 => \redbdir/libgnuastro.so.8 (0x00007febb419e000)\\
      \hspace{0.5cm}bz2.so.1.0 => \redbdir/libbz2.so.1.0 (0x00007febb3e20000)\\
      \hspace{0.5cm}\alert{m.so.6} => /usr/lib/libm.so.6 (0x00007febb4025000)\\
      \hspace{0.5cm}\alert{pthread.so.0} => /usr/lib/libpthread.so.0 (0x00007febb4004000)\\
      \hspace{0.5cm}\alert{c.so.6} => /usr/lib/libc.so.6 (0x00007febb3e3f000)\\
      \hspace{0.5cm}rt.so.1 => /usr/lib/librt.so.1 (0x00007febb3e35000)\\
      \hspace{0.5cm}dl.so.2 => /usr/lib/libdl.so.2 (0x00007febb3e1b000)\\
      \hspace{0.5cm}linux-vdso.so.1 (0x00007ffcf2497000)\\
      \hspace{0.5cm}/lib64/ld-linux-x86-64.so.2 => /usr/lib64/ld-linux-x86-64.so.2 (0x00007febb53c6000)
   }
  \end{frame}




  \begin{frame}{Advantages of this build system}
    \begin{columns}
      \column{7cm}
      \begin{itemize}
        \setlength\itemsep{1cm}
      \item No need for \alert{root}/administrator \alert{permissions}
        (on servers or super computers).
      \item Whole system is built \alert{automatically} on any
        Unix-like operating system (less 2 hours).
      \item Dependencies of different projects will \alert{not conflict}.
      \item (Almost) all depencies are \alert{exactly} documened and
        can be reproduced.
      \end{itemize}
      \column{4cm}
      \includegraphics[width=\linewidth]{img/unchained.jpg}\\
      \tiny \url{https://natemowry2.wordpress.com}
    \end{columns}
  \end{frame}



  \begin{frame}{Software acknowledgment and citation automatically generated in paper}
    \includegraphics[width=\linewidth]{img/software-cite.png}
  \end{frame}
  \begin{frame}{Software acknowledgment and citation automatically generated in paper}
    \includegraphics[width=\linewidth]{img/software-cite-highlighted.png}
  \end{frame}





%  \begin{frame}{Reproducing the result and report/paper}
%    The two \alert{simple} and \alert{familiar} commands below are
%    enough to exactly reproduce the results at any time.
%
%    \begin{itemize}
%    \item[] \texttt{\$ ./configure}
%    \item[] \texttt{\$ make}
%    \end{itemize}
%
%    With \texttt{./configure}, you specify the local directories to
%    use. All necessary \alert{software} are then \alert{downloaded}
%    and installed there (independent of your OS or other projects).
%
%    \vspace{0.3cm} With \texttt{make}, input \alert{data} from online
%    archives (databases) are \alert{downloaded}, if not locally
%    available, the processing is done, and the \LaTeX{} paper is built
%    as a PDF (e.g., see
%    \textcolor{blue}{\small\href{https://doi.org/10.5281/zenodo.1164774}{zenodo.1164774}}
%    or
%    \textcolor{blue}{\small\href{https://gitlab.com/makhlaghi/reproducible-paper-output/raw/master/paper.pdf}{template's
%        output}}).
%
%   \vspace{0.3cm} Enabling version control (e.g., with \alert{Git})
%    encourages testing different ideas while not harming the
%    initial/base result (thus encouraging \alert{creativity} and
%    brainstorming during the project).
%
%    \vspace{0.3cm} After publication, \alert{readers} can
%    \alert{change} the input configurations and the numbers and
%    figures of the reproduced paper will respectively change. This
%    encourages creativity and brainstorming after the project as well
%    as sharing of (the hardly gained) experiences with the whole
%    community.
%  \end{frame}



  \renewcommand{\nodeopacity}{1}
  \begin{frame}{Everything in plain text (machine and human readable)}
    \include{tex/project-graph} \end{frame}
  \newcommand{\gitlogo}{}
  \begin{frame}{Everything in plain text (machine and human readable)}
    \include{tex/project-graph}
  \end{frame}



  \begin{frame}{New projects branch from template} \include{tex/git-branch} \end{frame}
  \newcommand{\projbranch}{}
  \begin{frame}{New projects branch from template} \include{tex/git-branch} \end{frame}
  \newcommand{\mergebranch}{}
  \begin{frame}{New projects branch from template} \include{tex/git-branch} \end{frame}
  \newcommand{\tofuture}{}
  \begin{frame}{New projects branch from template} \include{tex/git-branch} \end{frame}
  \newcommand{\githappy}{}
  \begin{frame}{New projects branch from template} \include{tex/git-branch} \end{frame}


  \begin{frame}{Publication of the project}

    A reproducible project using this template will have the following
    (\alert{plain text}) components:
    \begin{itemize}
    \item Makefiles.
    \item \LaTeX{} source files.
    \item Configuration files for software used in analysis.
    \item Scripts/programming files (e.g., Python, Shell, AWK, C).
    \end{itemize}
    The \alert{volume} of the project's source will thus be
    \alert{negligible} compared to a single figure in a paper
    (usually $\sim100$ kilo-bytes).

    \vspace{1cm} The project's pipeline (customized template) can be
    \alert{published} in
    \begin{itemize}
    \item \alert{arXiv}: uploaded with the \TeX{} source to always
      stay with the paper \\(for example
      \textcolor{blue}{\small\href{https://arxiv.org/abs/1505.01664}{arXiv:1505.01664}}). The
      file containing all macros must also be uploaded so arXiv's
      server can easily build the \LaTeX{} source.
    \item \alert{Zenodo}: Along with all the input datasets (many
      Gigabytes) and software \\(for example
      \textcolor{blue}{\small\href{https://doi.org/10.5281/zenodo.1164774}{zenodo.1164774}}) and given a unique DOI.
    \end{itemize}
  \end{frame}



  \begin{frame}{GOOD NEWS: RDA adoption grant to IAC for this template}
      \begin{center}
        \includegraphics[width=3cm]{img/rda.png}\hspace{1cm}
        \includegraphics[width=1.8cm]{img/iac.png}

        \includegraphics[width=\linewidth]{img/h2020.png}
      \end{center}

      \vspace{1cm} For this template, the \alert{IAC} is selected as
      a \alert{Top European organization} funded to adopt RDA
      Recommendations and Outputs.

      \vspace{1cm}
      \scriptsize
      \begin{itemize}
      \item Research Data Alliance was launched by the \alert{European
        Commission}, NSF, National Institute of Standards and
        Technology, and the Australian Government’s Department of
        Innovation.
      \item RDA Outputs are the technical and social infrastructure
        solutions developed by RDA Working Groups or Interest
        Groups that enable data sharing, exchange, and
        interoperability.
      \end{itemize}

      \vspace{0.2cm}
      \centering

  \end{frame}



  \begin{frame}{Summary:}

    A fully working template/framework is introduced that will do the
    following steps/instructions (all in simple plain text files).
    \begin{itemize}
      \item \alert{Automatically downloads} the necessary
        \emph{software} and \emph{data}.
      \item \alert{Builds} the software in a \alert{closed
        environment}.
      \item Runs the software on data to \alert{generate} the final
        \alert{research results}.
      \item A modification in one part of the analysis will only
        result in re-doing that part, not the whole project.
      \item Using LaTeX macros, paper's figures, tables and numbers
        will be \alert{Automatically updated} after a change in
        analysis. Allowing the scientist to focus on the scientific
        interpretation.
      \item The whole project is under \alert{version control} (Git)
        to allow easy reversion to a previous state. This
        \alert{encourages tests/experimentation} in the analysis.
      \item The \alert{Git commit hash} of the project source, is
        \alert{printed} in the published paper and \alert{saved on
          output} data products.  Ensuring the
        integrity/reproducibility of the result.
    \end{itemize}

    \begin{tcolorbox}[width=\linewidth, boxsep=1pt, left=1pt, right=1pt,
                      top=1pt, bottom=1pt]
      For a technical description of the template's implementation, as
      well as a checklist to customize it, and tips on good practices,
      please see this page:

    \textcolor{blue}{\footnotesize\url{https://gitlab.com/makhlaghi/reproducible-paper/blob/master/README-hacking.md}}
    \end{tcolorbox}
  \end{frame}
\end{document}