Schedule
Wednesday, June 25th
Arrival in Brixen. If you plan to arrive before noon, you can join for the following social program (with the CSAMA group). You must let us know in advance as space is limited.
We meet in front of Palazzo Vescovile at 1:50 pm. We need to split into 3 groups (A, B, C)
Group A: guided tour in English starts at 2:00
Group B: guided tour in German starts at 2:15
Group C: guided tour in English starts at 3:00
Between 2pm and 3pm, participants of group C can take a free visit at other expositions in Palazzo Vescovile. Same option for groups A&B, from 3:00 to 4:00.
Due to bad weather forecast the programme below has been cancelled!
At 4:15pm we take the bus in front of the Office of Tourism (just behind Palazzo Vescovile), Viale Ratisbona 9 http://www.brixen.org. Bus will bring us to the Plose Ropeway. We will have a walk from the arrival of the ropeway to RossAlm.
We will have dinner at RossAlm (around 7pm) and then walk back (at 8:30 at most) to the bus waiting (but not forever) for us. Bus will leave from the parking lot at 9:30pm.
Thursday, June 26th
9:00-12:00
Closed meeting R Core/R Foundation members only. This will take place at the Casa della Gioventù, in the library.Session I: Extending R - performance enhancements, 1:00-3:15 main hall, Casa della Gioventù
| Luke Tierney | Some advances in the R Engine |
| Radford Neal | PQR and proposals for extending the R Language |
| Michael Haupt | FastR: Status and Outlook |
| Helena Kotthaus | Performance Analysis for R: Towards a Faster R Interpreter |
| Tomas Kalibera | Running R Faster |
| How R can be optimized for performance, based on my experimental changes to the R VM (the GNU-R implementation) | |
| Discussion | Panelists G. Sawitzki, Jan Vitek, Tony Rossini |
Session II: Compilation/Optimization, 3:45-5:30 main hall, Casa della Gioventù
| Alexander Bertram | IT Compilation in Renjin |
| Haichuan Wang | Optimizing R VM: Interpreter-level Specialization and Vectorization |
| Justin Talbot | Riposte: Just-in-time Length Specialization of Dynamic Vector Code |
| In this talk, I'll cover recent progress on the Riposte JIT compiler for R which has focused on understanding and improving the performance of code containing scalars and short vectors. I'll describe a trace-based just-in-time compilation strategy that performs partial length specialization, which is designed to avoid excessive compilation over-head while still enabling the generation of efcient machine code through length-based optimizations such as vector fusion, vector copy elimination, and the use of hardware SIMD units. | |
| Jan Vitek | Ideas for future implementations of R |
| Experience with our Java based implementation and our thoughts for a new implementation that would be better integrated with the current one and would use LLVM as a backend. | |
| Discussion | Luke Tierney, Radford Neal, Simon Urbanek |
Friday, June 27th
Discussion Session, 9:00-12:00 library
Discussion to consider setting up a "Consortium" of R users and interested companies (Revolution, RStudio, Google, others ... perhaps including Genentech) under the auspices of the Linux Foundation. It would be a 510(c)(6) ("trade association") nonprofit, funded by dues from its corporate members. Its mission would be to promote the R Project and to support the R Community through projects selected by its members and the community. The next-generation R Forge project would be the first project, and we have ideas for others. Setting up this consortium through the Linux Foundation eliminates legal complexities and gives us access the Foundation's technical, IT, marketing, events and legal resources.Session III, 1:00-3:15 main hall, Casa della Gioventù
| Simon Urbanek | RCloud - Integrating Exploratory Visualization, Analysis and Deployment |
| RCloud is a collaborative exploratory data analysis platform designed for teams of researchers to share both results and methodology. Instead of first performing analysis, then developing an application to present the results, the process of exploration and sharing results is integrated into a single workflow. Every RCloud notebook is already deployed as a web application when it is created, integrating JavaScript visualization tightly with R data analysis. By default, scripts, data feeds, and reports are public in RCloud, and RCloud includes an integrated search feature to help find others' results. In this talk we introduce and discuss RCloud as well as its integration into distributed computing frameworks such as Hadoop. | |
| JJ Allaire | Packrat - A Dependency Management System for R |
| Michael Lawrence | GRAN a package management system |
| David Smith | A proposal for a distribution of R focused on reproducibility |
| Deepayan Sarkar | A new graphics API |
| Discussion | Bill Dunlap, Antony Unwin, Kurt Hornik |
Session IV, 3:45-5:30 main hall, Casa della Gioventù
| Vincent Carey | Managing very large-scale testing procedures with R |
| Joe Cheng | Reactive programming in R |
| Romain Francois | Rcpp11 |
| Lee Edlefsen | Parallel External Memory Algorithms Using R Reference Classes |
| This is about a framework we are developing enabling code to be written in R, using reference classes, that can be executed on data that is too large to fit into memory and can also executed in parallel on one or more computers, resulting if desired in a single unified result. We will release the underlying framework as open source, and code written this way can also be be executed on the distributed computing platforms Revolution supports, such as IBM LSF, Windows HPC Server, Hadoop, and Teradata. | |
| Radford Neal | Proposals for Extending the R Language |
| Discussion | Duncan Murdoch, Jan Vitek, Karl Miller |