Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
To R or Not to R? (undirectedgrad.blogspot.com)
31 points by fogus on Aug 18, 2009 | hide | past | favorite | 17 comments


http://incanter.org/

Incanter is a "Clojure-based, R-like statistical computing and graphics environment for the JVM". Ready to replace R? I don't know, R has a lot of libraries. Worth a look? Definitely.


Colt (Parallel Colt, really) is what Incanter is built upon and it already has a lot of libraries. I suppose R has more, though.

However, there is no Incanter community yet - not even a mailing list. That matters.


OK, now there is a mailing list. http://groups.google.com/group/incanter/


One certainly could write an adapter layer to the R libs. I think the interfaces are pretty coarse-grained from what I recall.


I wonder why he isn't considering Octave, given that his lab already has a pile of existing Matlab code. Also, I think it's relatively easy to get Python to go faster than Matlab: http://www.scipy.org/PerformancePython so maybe that's a reasonable approach if some people are already using it but just having trouble making it go fast.


I can only confirm that R has a great community of savy-users ready to help. See the r-help mailing list.


I tried to use R for a project, and I found the small size of the community and available documentation to be frustrating. In addition, R suffers from the "curse of C" , where it is near impossible to google for help on a language with a one-letter name. Though, perhaps I just managed to miss that particular mailing list in my search for R materials.

In contrast, Matlab has excellent resources available for the beginner.


Try http://rseek.org - a Google custom search engine for R materials.


Small size of the community? In comparison to what?


I'm in exactly the same situation. MATLAB and R are so different, yet there's no obvious way to distinguish which will be better for the job.

Thanks to those who are also proposing alternatives. However, if the author is in anything of the same situation that I am, then it must be one of the two. Anything that's not "tried and true" seems to make too many people (bosses, clients) squeamish.


So, what was the problem with Matlab that made you want to switch to R? I didn't see anything in the article about it...

Also, there are some good machine learning libraries for .NET, which you could call from C# or F#. If you don't need all the other features of Matlab or R (or you want some of the .NET features), maybe you consider that as an option.

EDIT: Also...there are plugins for Matlab (both commercial and free) that will use CUDA to accelerate the calculations. Depending on what you're doing, it could be a big boost to your research, so check that out as well.


"free (very important!) if we want to run a job on 100 machines (e.g. in the cloud), I believe currently you need a matlab licence for each one"


No-one pays for high-end commercial software like that; you find a salesman near the end of the financial year when he wants to make his numbers and he cuts you a deal on a site license.

With computation tho' time very literally is money. If you can get the job done with commercial product X in the same time with less kit than it would take to do with free product Y, the point at which it becomes worthwhile to do so - if at all - is a no-brainer.


Easier said than done; besides why deal with complicated proprietary licenses when an open one will do? You have to account for future upgrades too (they may not get you now but they'll nail you hard later) Plus it doesn't sound like the guy who wrote the post is shipping a product ( which is when many open licenses start giving some trouble).

I do agree if everything overall adds up in favor of the proprietary solution, then they should just use it; but in many cases I think the open product will have the long term edge especially in this economy


Wish I had the time to try the same task in R, NumPy/SciPy and Incanter to get a feel for the relative strengths and weaknesses of each. If anyone does have the time, and cares to write it up in a blog, I would be happy to read the results. :)


tor V ~tor => T


Yes, the proposition is always true, but that doesn't help him to make a decision.

At work, some people prefer R; others Matlab. The nice thing bout R is that it's free so you can have it on multiple machines.

But I'm wondering why he hasn't considered other alternatives like octave, etc.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: