High-Performance Computing in R
The meeting will take place on the 7th floor in a room provided by Level7.
R is an incredibly rich and powerful language for exploratory data analysis and scientific computing. But, often mistakenly, R is not considered a good language for high-performance computing. In fact the language provides a remarkably comprehensive set of tools for high-performance computation.
We'll discuss tips and tricks for writing R code for high-performance computation, including:
- Using fast low-level numeric libraries
- Writing vectorized code
- Working with data larger than memory
- Parallel processing with 'parallel,' 'foreach,' and others
- Using the right algorithms
We'll focus on practical computation and show examples of fast large-scale principal components and generalized linear models among others. Feel free to come with questions about performance!
Bryan Lewis has worked with R for many years and is the author of several R packages including irlba, rredis, doRedis, websockets, bigalgebra, scidb, and others. He is the chief data scientist at Paradigm4 in Waltham, MA and has a Ph.D. in applied mathematics.