Microsoft R Server 9.1 now available
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.
During today's Data Amp online event, Joseph Sirosh announced the new Microsoft R Server 9.1, which is available for customers now. In addition the updated Microsoft R Client, which has the same capabilities for local use, is available free for everyone on both Windows and — new to this update — Linux.
This release adds many new capabilities to Microsoft R, including:
Based on R 3.3.3. This update is based on the latest release of open source R from the R Foundation, bringing many improvements to the core language engine.
Microsoft R Client now available on Linux. The Microsoft R Client is available free, and has all of the same capabilities as Microsoft R Server for a standalone machine (or, you can use it to push heavy workloads to a remote Microsoft R Server, SQL Server or compute cluster). With this release it is now also available on Linux (specifically, Ubuntu, Red Hat / CentOS and SUSE) in addition to Windows.
New function for “pleasingly parallel” R computations on data partitions. The new rxExecBy function allows you to apply any R function to partitions of a data set, and compute on the partitions in parallel. (You don't have to manually split the data first; this happens automatically at the data source, without the need to move or copy the data.) You can use any parallel platform supported in Microsoft R: multiple cores in a local context, multiple threads in SQL Server, or multiple nodes in a Spark cluster.
New functions for sentiment scoring and image featurization have been added to the built-in MicrosoftML package. The new getSentiment function returns a sentiment score (on a scale from “very negative” to “very positive”) for a piece of English text. (You can also featurize text in English, French, German, Dutch, Italian, Spanish and Japanese as you build your own models.) The new featurizeImage function decomposes an image into a few numeric variables (using a selection of ResNet recognizers) which can then be used as the basis of a predictive model. Both functions are based on deep neural network models generated from thousands of compute-hours of training by Microsoft Research.
Interoperability between Microsoft R Server and sparklyr. You can now use RStudio's sparklyr package in tandem with Microsoft R Server in a single Spark session
New machine learning models in Hadoop and Spark. The new machine learning functions introduced with Version 9.0 (such as FastRank gradient-boosted trees and GPU-accelerated deep neural networks) are now available in the Hadoop and Spark contexts in addition to standalone servers and within SQL Server.
Production-scale deployment of R models. The new publishService function creates a real-time web-service for certain RevoScaleR and MicrosoftML functions that is independent of any underlying R interpreter, and can return results with millisecond latency. There are also new tools for launching and managing asynchronous batch R jobs.
Updates for R in SQL Server. There are also improvements for R Services in SQL Server 2016, in addition to the new R functions described above. There are new tools to manage installed R packages on SQL Server, and faster single-row scoring. Even more new capabilities are coming in SQL Server 2017 (now in preview for both Windows and Linux), including the ability to use both R and Python for in-database computations.
For more on the updates in Microsoft R Server 9.1, check the official blog post by Nagesh Pabbisetty linked below.
SQL Server Blog: Introducing Microsoft R Server 9.1 release
R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.