The Julia Programming Language in 2020 (for Data Science)

Subscribe to RichardOnData here:    / @richardondata  

See my follow-up video on top Julia packages here:    • 10 Julia Packages You Should Learn for Dat...  

Learn more about Julia's speed here (credit to Sayan Sinha over at HackerNoon): https://hackernoon.com/performance-an...

In this video I discuss the Julia programming language, as it stands in 2020, with a focus on data science as it compares to the tried and true juggernauts R and Python. Julia was developed in 2012 by Alan Edelman, Jeff Bezanson, Stefan Karpinski, Deepak Vinchhi, Keno Fischer, and Viral Shah. They brought together some brilliant mathematical minds with focuses on combining the ease of use, utility, and syntax of Python with the performance of C. It is open-source with many packages developed for data science.

From a computational standpoint, Julia is built for multiple dispatch, enables asynchronous I/O, and it's compiled rather than interpreted. These are all benefits that enable flexibility and high performance.
Julia is very quick and easy to learn, in much the same way Python is.
From a runtime perspective, Julia runs faster than Python. This was demonstrated by the Julia Lab themselves; however third parties back this up with larger experiments.
There is one current caveat with speed called the "time to first plot" problem, but it is being worked on by developers.
The DataFrames.jl package helps it work great for data wrangling & manipulation.
For visualization you have many options for like Plots.jl, Gadfly.jl, and VegaLite.jl. I don't like any of these as much as ggplot2 in R but they get the job done.
No answer to RShiny yet.
The IDEs are Juno, or you can download iJulia to interface with Jupyter Notebooks.
Many options for statistical modeling and machine learning, including the ScikitLearn.jl interpretation (or use PyCall to directly use it from Python). StatsModels.jl, MultivariateStats.jl, and Distributions.jl are some examples of statistical packages.
A full deep learning framework through Flux.jl is available.

Overall, it is difficult for Julia to compete with some of R and Python's capabilities, but you have to remember that speed and runtime is of the essence, and we live in an era of very big data. Microseconds can make a big difference and Julia may be able to deal with enormous data in a way that is simply not practical for R or Python. For this reason I think it could be a huge player in 3-5 years time given some maturity and development of its user community.

#JuliaForDataScience #DataScience #JuliaProgramming

PayPal: richardondata@gmail.com
Patreon:   / richardondata  
BTC: 3LM5d1vibhp1F7pcxAFX8Ys1DM6XLUoNVL
ETH: 0x3CfC599C4c1040963B644780a0E62d45999bE9D8
LTC: MH8yPjvSmKvpmRRmufofjRB9hnRAFHfx32