Introduction to Julia for Pythonistas

<img src="http://pearsonlab.github.io/images/plab_hex_icon_gray.png" width="150", align="left", style="float:left"> <img src="http://www.duke.edu/~jmp33/assets/dibs-logo-small.png" width="350", align="left", style="float:right">

John Pearson
P[λ]ab
Duke Institute for Brain Sciences

Following along

If you have Julia...

Slides and other material available at https://github.com/jmxpearson/julia-for-pythonistas

About me

• @jmxpearson on GitHub and (rarely) Twitter; pearsonlab.github.io
• computational neuroscience lab at Duke
• using Julia for about 1.5 years
• member of JuliaStats organization

What is Julia?

• new programming language (public since 2012)
• brought to you by MIT, Julia Computing
• focused on scientific computing applications
• aims to be a high-level prototyping language that is also fast enough for serious simulation
  • as easy to learn and flexible as Python or R
  • within 2x the speed of C, C++, or FORTRAN
• Currently (almost) 0.5, with very active developer community

Reasons to Consider Julia

• you mostly write your own algorithms
• you need to write code that is very fast
• you’d like to prototype and optimize within the same language
• you want zero overhead boilerplate calls to C, Python (and soon C++)
• you need powerful metaprogramming/macros
• you want access to state-of-the-art automatic differentiation and optimization
• you want to use Greek letters in variable names
• you are interested in programming language design
• you want to get in on the ground floor

Reasons Julia may not be for you (yet)

• you mostly use packages written by others
• you rely critically on several packages that aren't available
• you don’t have much need to optimize your code
• you perform mostly data analysis and model building, as opposed to algorithm development
• you prioritize long-term stability
• you don’t want to devote time to keeping up with changes
• you have zero background with compiled languages