ENTRIES TAGGED "R"
Data Science tools: Are you “all in” or do you “mix and match”?
It helps to reduce context-switching during long data science workflows.
An integrated data stack boosts productivity
As I noted in my previous post, Python programmers willing to go “all in”, have Python tools to cover most of data science. Lest I be accused of oversimplification, a Python programmer still needs to commit to learning a non-trivial set of tools1. I suspect that once they invest the time to learn the Python data stack, they tend to stick with it unless they absolutely have to use something else. But being able to stick with the same programming language and environment is a definite productivity boost. It requires less “setup time” in order to explore data using different techniques (viz, stats, ML).
Multiple tools and languages can impede reproducibility and flow
On the other end of the spectrum are data scientists who mix and match tools, and use packages and frameworks from several languages. Depending on the task, data scientists can avail of tools that are scalable, performant, require less2 code, and contain a lot of features. On the other hand this approach requires a lot more context-switching, and extra effort is needed to annotate long workflows. Failure to document things properly makes it tough to reproduce3 analysis projects, and impedes knowledge transfer4 within a team of data scientists. Frequent context-switching also makes it more difficult to be in a state of flow, as one has to think about implementation/package details instead of exploring data. It can be harder to discover interesting stories with your data, if you’re constantly having to think about what you’re doing. (It’s still possible, you just have to concentrate a bit harder.)
MATLAB, R, and Julia: Languages for data analysis
Inside core features of specialized data analysis languages.
Big data frameworks like Hadoop have received a lot of attention recently, and with good reason: when you have terabytes of data to work with — and these days, who doesn’t? — it’s amazing to have affordable, reliable and ubiquitous tools that allow you to spread a computation over tens or hundreds of CPUs on commodity hardware. The dirty truth is, though, that many analysts and scientists spend as much time or more working with mere megabytes or gigabytes of data: a small sample pulled from a larger set, or the aggregated results of a Hadoop job, or just a dataset that isn’t all that big (like, say, all of Wikipedia, which can be squeezed into a few gigs without too much trouble).
At this scale, you don’t need a fancy distributed framework. You can just load the data into memory and explore it interactively in your favorite scripting language. Or, maybe, a different scripting language: data analysis is one of the few domains where special-purpose languages are very commonly used. Although in many respects these are similar to other dynamic languages like Ruby or Javascript, these languages have syntax and built-in data structures that make common data analysis tasks both faster and more concise. This article will briefly cover some of these core features for two languages that have been popular for decades — MATLAB and R — and another, Julia, that was just announced this year.
MATLAB
MATLAB is one of the oldest programming languages designed specifically for data analysis, and it is still extremely popular today. MATLAB was conceived in the late ’70s as a simple scripting language wrapped around the FORTRAN libraries LINPACK and EISPACK, which at the time were the best way to efficiently work with large matrices of data — as they arguably still are, through their successor LAPACK. These libraries, and thus MATLAB, were solely concerned with one data type: the matrix, a two-dimensional array of numbers.
This may seem very limiting, but in fact, a very wide range of scientific and data-analysis problems can be represented as matrix problems, and often very efficiently. Image processing, for example, is an obvious fit for the 2D data structure; less obvious, perhaps, is that a directed graph (like Twitter’s follow graph, or the graph of all links on the web) can be expressed as an adjacency matrix, and that graph algorithms like Google’s PageRank can be easily implemented as a series of additions and multiplications of these matrices. Similarly, the winning entry to the Netflix Prize recommendation challenge relied, in part, on a matrix representation of everyone’s movie ratings (you can imagine every row representing a Netflix user, every column a movie, and every entry in the matrix a rating), and in particular on an operation called Singular Value Decomposition, one of those original LINPACK matrix routines that MATLAB was designed to make easy to use.
Big Data: SSD’s, R, and Linked Data Streams
If you haven’t seen it, I recommend you watch Andy Bechtolsheim’s keynote at the recent Mysqlconf. We covered SSD’s in our just published report on Big Data management technologies. Since then, we’ve gotten additional signals from our network of alpha geeks and our interest in them remains high. I had a chance to visit with Dataspora founder and blogger Mike Driscoll, an enthusiastic advocate for the use of the open source statistical computing language, R.
