5 ms·

Python Overtakes R in Data Science and Machine Learning Platforms

38 points by sus_007 9y ago

didgeoridoo 9y ago
Genuine question: is there a reason to use R over Python? R was actually my first programming language back in high school stats, but I never took to it the way I did with Python, Ruby, etc. Maybe because we never really learned the fundamentals and kind of bodged our code together?
- davidwihl 9y ago
  For a statistician, R has a variety of packages that are unavailable in Python. RStudio is a very easy to use cross platform dev environment. With a different R kernel, you can take advantage of multiple cores.
  - cluoma 9y ago
    And the huge variety of presentation tools such as ggplot2, Rmarkdown, and Shiny still makes R a very attractive choice. Tight integration with Rcpp also makes it very easy to speed up calculations when performance becomes an issue.
    - type0 9y ago
      For ggplot2 there is plotnine https://github.com/has2k1/plotnine https://github.com/has2k1/plotnine (disscussion https://news.ycombinator.com/item?id=14430447 https://news.ycombinator.com/item?id=14430447) Also Python is not a "statistician" language so the adoption there is very slow.
    - deskamess 9y ago
      For me Rmarkdown is such a win for R. Web deployment, especially the web servers to host shiny apps is a weak point for R (Linux specific, and really not many options even in the Linux arena). Then again, perhaps notebooks are a better way to handle online data exploration.
  - nthcolumn 9y ago
    Which packages would you say are the most glaring omissions? I thought that R would become even more popular now it is on Azure ML - (if you haven't checked MSFT Azure ML out you really should - it's a blast). I use both (cut-paste monkey!). There seems to be a bit more magic with R, less code. Shiny is pretty cool too. btw - shinyproxy package was compromised recently - check your download hashes :(
    - davidwihl 9y ago
      The most glaring package omissions are related to Machine Learning, e.g. 1) an R equivalent of sci-kit learn 2) A native R equivalent than the buggy interface to TensorFlow 3) A higher level ML equivalent to Keras or PyTorch.
- ploika 9y ago
  For data analysis and a lot of (maybe most) machine learning tasks, not really. For statistical analysis and modelling, R is still preferable to Python in my experience.
go_prodev 9y ago
I use both languages daily, for different purposes. R is incredible for data analysis and visualisation. I find it just blends into the background and let's me focus on the data. Python is so versatile and pleasant to work with. But for data analysis it is a bit more verbose and takes some focus away from the data.

38 points by sus_007 9y ago

Genuine question: is there a reason to use R over Python? R was actually my first programming language back in high school stats, but I never took to it the way I did with Python, Ruby, etc. Maybe because we never really learned the fundamentals and kind of bodged our code together?

davidwihl 9y ago
For a statistician, R has a variety of packages that are unavailable in Python. RStudio is a very easy to use cross platform dev environment. With a different R kernel, you can take advantage of multiple cores.
- cluoma 9y ago
  And the huge variety of presentation tools such as ggplot2, Rmarkdown, and Shiny still makes R a very attractive choice. Tight integration with Rcpp also makes it very easy to speed up calculations when performance becomes an issue.
  - type0 9y ago
    For ggplot2 there is plotnine https://github.com/has2k1/plotnine https://github.com/has2k1/plotnine (disscussion https://news.ycombinator.com/item?id=14430447 https://news.ycombinator.com/item?id=14430447) Also Python is not a "statistician" language so the adoption there is very slow.
  - deskamess 9y ago
    For me Rmarkdown is such a win for R. Web deployment, especially the web servers to host shiny apps is a weak point for R (Linux specific, and really not many options even in the Linux arena). Then again, perhaps notebooks are a better way to handle online data exploration.
- nthcolumn 9y ago
  Which packages would you say are the most glaring omissions? I thought that R would become even more popular now it is on Azure ML - (if you haven't checked MSFT Azure ML out you really should - it's a blast). I use both (cut-paste monkey!). There seems to be a bit more magic with R, less code. Shiny is pretty cool too. btw - shinyproxy package was compromised recently - check your download hashes :(
  - davidwihl 9y ago
    The most glaring package omissions are related to Machine Learning, e.g. 1) an R equivalent of sci-kit learn 2) A native R equivalent than the buggy interface to TensorFlow 3) A higher level ML equivalent to Keras or PyTorch.
ploika 9y ago
For data analysis and a lot of (maybe most) machine learning tasks, not really. For statistical analysis and modelling, R is still preferable to Python in my experience.

go_prodev 9y ago

I use both languages daily, for different purposes. R is incredible for data analysis and visualisation. I find it just blends into the background and let's me focus on the data. Python is so versatile and pleasant to work with. But for data analysis it is a bit more verbose and takes some focus away from the data.