Computational Science in Julia - An introduction course into the Julia programming language

--- title: 'Computational Science in Julia - An introduction course into the Julia programming language ' tags: - Julia - course - beginners - plotting - data analysis - agent based model authors: - name: Andreas Kuhn orcid: 0000-0001-8146-2095 affiliation: 1 - name: Sabine C. Fischer orchid: 0000-0001-7237-814X affiliation: 1 affiliations: - name: Center for Computational and Theoretical Biology , University of Würzburg index: 1 date: 23/01/23 bibliography: paper.bib pandoc_args: ["--toc", "--toc-depth=2",] output: pdf_document --- # Computational Science in Julia - An introduction course into the Julia programming language # Scientific programming in Julia - An introductory course ## Summary This course provides an introduction into scientific programming in ``Julia``. Compared to other popular languages in this field like ``Python`` or ``C++``,``Julia`` has been specifically designed for scientifc applications. hier fehlen noch ein, zwei sätze No previous knowledge is necessary to follow the nine course lectures. Every lecture contains one Jupyter Notebook, where executable code examples are shown together with markdown cells containing specific explanations about the shown code but also general concepts about programming in ``Julia``. The first five lectures cover the basic language syntax (e.g. data types, package installation, functions, ...). The last four lectures are targeted towards scientific applications and cover topics like plotting, data import/export, data analysis and a simple agent-based simulation. Each lecture ends with exercises, which directly apply the learned concepts. Sample solutions for all exercises are provided. After the course is finished, one is equipped with all the necessary tools to start scientific programming in ``Julia``. ## Statement of Need The ``Julia`` programming language first appeared in 2012 [@bezanson_julia_2012] and had two main design goal from the start: Speed and interactiviy which makes it a perfect fit for scientific computation and data analysis. Since then, it has seen a rapidly growing adaptation in various fields of science. Ranging from black holes [@fernandes_new_2022] to quantum systems [@gawron_quantuminformationjljulia_2018; @kramer_quantumopticsjl_2018] in physics to applications in biology [@roesch_julia_2021; @knopp_mrirecojl_2021], economics [@coleman_matlab_2021], machine learning [@gao_julia_2020] and many more. This fast acceptance of a new programming language is quite remarkable, as programming languages are a field that is historically known to be very slow at pivoting (e.g. ``Python`` and ``C++`` are both over 30 years old). One reason for this is that ``Julia`` solves the two language problem [@perkel_julia_2019] that has been tormenting many researcher for decades: For example, ``Python`` or ``Matlab`` are used for fast prototyping and plotting, ``R`` for statistical analyses and a compiled language like ``C/++`` or ``Fortran`` for computationally expensive simulations. Using two or more languages creates a lot of unnecessary overhead: Firstly, researcher have to spend more time learning to code in two languages (which is especially true for ``C/++`` or ``Fortran``) and secondly, exchanging data and logic between languages can be very tedious and time-consuming. Julia solves this problem, as all the mentioned tasks can be accomplished in Julia alone. Its simple syntax is heavily inspired by Python and it features , which allows fast prototyping [@andreaskuhn_how_2022]. The plotting package ``Makie`` [@DanischKrumbiegel2021] provides similar capabilities as ``matplotlib`` in ``Python`` or ``ggplot2`` in ``R``. The meta package ``StatsKit`` includes all necessary tools to do sophisticated statistical analysis. And most importantly, ``Julia`` is fast. Depending on the benchmark, ``Julia`` is on pair or within a factor of 2 to ``C`` [@al_julia_nodate; @churavy_bridging_2022]. ![](https://i.imgur.com/Fi8Acb1.png) Therefore, there is no better time than now to give Julia and its package ecosystem a try. Eventhough there are already free Julia courses available online ([1](https://carpentries-incubator.github.io/julia-novice/),[2](https://www.datacamp.com/courses/introduction-to-julia),[3](https://juliaacademy.com/courses)), we believe, that this course offers additional value for (new) scientific programmers. As it does not require any prior knowledge, is self-contained and build on one another, introduces many relevant topics in scientifc programming and offers training exercises together with sample solutions to each lesson. ## Details of the course ### Learning objectives The learning resource aims to enable learners to: * use interactive notebooks * understand the basic structure of a programming language * use Julia in a scientifc context: * import/export data * analyse data * plot data ### Content 1. Jupyter Notebooks 2. Package Installation 3. Datatypes 4. Conditional Iterators 5. Functions 6. Plotting 7. Import/Export 8. Data Analysis 9. Application: Simple Random Walk Simulation ### Instructional design The Jupyter notebooks contain four different types of cells: - General explanations - Executable code examples - Exercises - Meta comments/hints/notes The first three cell types are essential in understanding the contents of the course. The exercises are divided into three difficulties: easy, medium and hard. Easy and medium exercises can be solved using the content of the course alone, whereas some hard exercises require additional research, to transfer the acquired knowledge to new problems. The "meta" cells offer additional but non essential information. These, can be comments about the differences of Julia compared to other languages, advice on good coding practices,... . Whereas first time programmers can skip these cells if they want to, we expect the meta cells to be particularly useful for already experienced programmers switching to ``Julia``. ### Target audicence The main target audience of this course are scientists who want or are already using programming in their work. But no prerequisites are needed, therefore anybody can follow the course without problems. Programmers who already have experience in other languages should be able to finish the first five lession quite fast, as a lot of gerneral programming concepts are introduced there as well. But the last four more `Julia` specifc lessons should be very useful for them nevertherless. ## Experience of use We started using Julia in our own work out of frustation of the execution speed of raw Python. The choice of Julia compared to other compiled and fast languages like C/++, Fortran or Rust was easy, as Julia allowed us to keep the familar and convenient interactive development workflow that Python estabilshed. During the switch, we also had to learn Julia from ground up. And what is the best way to learn something new? Right, teaching it. Therefore, we developed this course with the original purpose to educate ourselves and new members of our group on Julia. Since the begining of 2022, seven other people have completed the material as a self-study course. Furthermore, we have been providing the course material on our website as a free [download](https://www.biozentrum.uni-wuerzburg.de/cctb/research/supramolecular-and-cellular-simulations/) since March 2022. The feedback was used to iteratively improve the course over time. We believe that the course has reached a level of maturity and that it offers great value to everybody who wants to learn Julia. Next, we plan to host a training workshop in a research consortium. In addition, we are convinced that the course could be readily taught in a weekly manner as part of an undergraduate or graduate programme. Therefore, we invite everybody to use the course independently to learn or to teach Julia. ## Get started -> later only in readme To run the course material on your local machine, clone this [github repository]( https://github.com/AndreasKuhn-ak/WS2022_Julia). The repository contains a file called "Julia_course_manual.pdf" which lists the required software and how to install it. For a first experience without any prior installation, you can also use [mybinder](https://mybinder.org/v2/gh/AndreasKuhn-ak/WS2022_Julia/HEAD) to directly execute the Jupyter nootebooks in your browser. Unfortunately this comes with some restrictions: Due the limited server capacities of mybinder, the container building phase can sometimes fail and the GPU accelerated plotting in the ninth part will not work. You should also keep in mind that your changes will only be saved as long as your browser session is active, but you can download the changed notebooks to your local machine. ## Figures Figures can be included like this: ![Example figure.](figure.png) ## To Dos: * https://openjournals.readthedocs.io/en/jose/review_checklist.html Wir müssen sicher gehen, dass wir diese Aspekte alle abdecken * target audience * https://jose.theoj.org/papers/10.21105/jose.00156 Artikel mit sehr ähnlichem Thema * github repo mit documentation die das hier beinhaltet: https://jose.theoj.org/papers/10.21105/jose.00156 * Kurs Korretur lesen, damit keine (möglichst wenige) Typos drin sind * sie betonen überall immer, dass sie gerne die Geschichte des Kurses im ARtikel lesen möchten - das sollte also irgendwo mit drin sein * adaptation from others * comunity notes: Anpassungen von anderen -> readme ## Acknowledgements We acknowledge contributions from Simon Schardt, Christopher Nauroth-Kreß, Alexnder Leipold and the various students who served as beta testers for this course. ## Comments: ## References