While reading articles about pros and cons of programming languages for statisticians and researchers I never came across anyone, who would suggest Functional Programming instead of Octave or R and I'm trying to understand myself, whether it's because of Octave/R shine so much in the field of quickly prototyping some algorithms or maybe those, who compare Octave, R and Julia simply don't have any background in other types of programming.
Even Professor Andrew Ng from Stanford University (Machine Learning Class) suggests prototyping in Octave and then re-implementing in more general purpose programming languages such as C/C++ or Java for better performance. But how about using let's say Scala programming language for both prototyping and production implementations?
I'm asking this question, because I use Scala myself and I find it to be very good language for both prototyping and production ready systems. Most of the programming assignments for Probabilistic Graphical Models Course given by Professor Daphne Koller, I believe it would be much easier and faster to develop in Scala rather than in Octave, with the goodness of map, flatMap, reduce, prod and sum Scala functions, and an ability to use object oriented programming for data representation instead of putting everything into Octave matrices. (Scala is both object oriented and functional programming language).
I would like to hear some opinions here from someone with an experience in both Scala (or other functional/object oriented programming language) and R/Octave. I'm considering learning R myself or simply prototyping in Octave instead of still using Scala but I'm trying to find a good justification for it.
For this question I'm asking about the pure programming language and not the availability of some stats packages where R rocks.
Thank you for any comments.
[link] [30 comments]