Hi everyone, I know there are a few Python Programmers here and a few R coders. Mid last year R Studio released a beta version of "Reticulate" which allows for integration between Python and R. The package has been released on CRAN and it is very exciting to say the least. Both programs at their heart are based on C++. This is what Reticulate uses to communicate between the 2 languages. As much as I love R, Pyhton is much better working with MongoDB which is the reason why I am starting to use it. Maybe some of you will find it useful as well!!! https://cran.r-project.org/web/packages/reticulate/reticulate.pdf https://rstudio.github.io/reticulate/index.html https://resources.rstudio.com/webinars/r-rstudio-1-2-amp-python-a-love-story-sean-lopp
Big python fan here, use it exclusively for my work with pytorch and tensorflow. I never liked R at all, many packages I came across contain errors and are very poorly peer reviewed, especially financial math related packages. That R over the past years was bestowed with links to C/C++, C#, now python only shows that it never was a good research language to stand on its own. Only since datastructures via C/C++ ports were introduced did it get faster. I remember when R was single threaded, 32bit and as slow as a crawl.
I am actually a fan of Java and C#. I read a few books on python and found its library more readily usable, but personally prefer strong typing. I guess should familiarise myself with python for data testing
Have you checked out Microsoft's "new", free ML library for C#, etc? https://en.wikipedia.org/wiki/ML.NET
Python 3 has a very good compromise with type hints. I use this with mypy to catch many, but not all, errors.
Production for what exactly? I use a self written C# wrapper around the C++ api of tensorflow in production to use trained deep learning models to make predictions in my trading algorithms. Python functions as development and profiling tool for which it was mainly designed. Very few production level end products rely entirely on python, an example being Dropbox.
"Production" is the term used in the engineering industry for product manufacturing. I also prefer strong typing languages. Probably because I started with those. Python is relatively a new language.
Depends on the application. I write PySpark pipelines for big data all the time and deploy models in 50 lines of code or less. You don't have to write much code for model deployment any more, as much of the work has been subsumed by high-level ML APIs or products such as DataRobot. https://developers.google.com/machine-learning/guides/rules-of-ml/ Regards, PTR