by Joseph Rickert
Tracking R user group meetings is a good way to stay informed about what's happening in the R world. On Tuesday the Bay Area useR Group (BARUG) met at AdRoll in San Francisco. It was a mini-conference with 6 talks:
- Bryan Galvin our host at AdRoll (many thanks for the pizza and beer) kicked off the evening by showing how his company has created Sparkle, a big-league workflow for deploying Shiny Apps.
- Shivaram Venkataraman of Amplab gave an overview of SparkR, highlighting new features and giving a preview of what's coming.
- Perry de Valpine presented NIMBLE, a new C++ compiled language for building BUGS models with R that Bayesian R users should find of great interest.
- Hamed Dagour presented three scenarios's of how Personal Capital uses R for financial applications.
- Ramnath Vaidyanathan delivered a high energy, overview of htmlwidgets for R. Html widgets may very well be "the next big thing".
- Bill Grosso finished up the evening by describing how his team at Scientific Revenue use R to build dynamic pricing models for internet game companies.
SparkR has been generating quite a bit of interest in the R community at large. Shivaram revealed that SparkR is now going mainstream within the Spark project. SparkR will be merged into Spark in the upcoming 1.4 release. This means that there will be a Spark API for R.
The following slide illustrates the data frame methods that have been developed for SparkR:
Future SparkR work which is highlighted in Shivaram's presentation includes:
- High-level APIs for machine learning algorithms
- Pipelines with "featurizers"
- Extended models and summary methods
- APIs for streaming and time series analysis
- Distributed matrix operations
Ramnath's presentation highlighted some brand new work in html widgets. Really notable was his live demonstration of cooperating html widgets. Using message passing to transmit state status, Ramnath showed a cluster of graphs changing on the fly to reflect new information from graphs that they were linked to. This gif link gives an idea of the kinds of cool things people are doing with htmlwidgets.
In other R user group meetings around the world this week: The New Delhi UseR Group gave R tutorials, the Taiwan R User Group explored the MLDM package and its applications. Statistical Programming DC looked into R and Shiny. MadR looked into making R code easier and more reliable.The Portland R User Group got a glimpse of new features coming to RStudio. The Barcelona R users group will look into dplyr. Dublin R will look at using R with medical statistics method comparison studies. DataPhilly will investigate data science in health care and look into using R in Plotly. The Denver R Users Group will discuss websites with GitHub and RStudio in the context of continuous delivery systems. The Atlanta R Users Group will look into building a netter data stack.
It is only a small sample of 15 talks, but they do emphasize that both Shiny and the use of R in production level work are both trending topics in the R Community. Bryan's talk illustrates the practical intersection of both trends.