Skip to main content

Parallel Computing in R

This course covers in detail the tools available in R for parallel computing.

Start Course for Free
4 Hours15 Videos53 Exercises10,162 Learners4250 XP

Create Your Free Account

GoogleLinkedInFacebook

or

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.

Loved by learners at thousands of companies

Course Description

With an increasing amount of data and more complex algorithms available to scientists and practitioners today, parallel processing is almost always a must, and in fact, is expected in packages implementing time-consuming methods. This course introduces you to concepts and tools available in R for parallel computing and provides solutions to a few important non-trivial issues in parallel processing like reproducibility, generating random numbers and load balancing.
  1. 1

    Can I Run My Application in Parallel?

    Free

    In order to take advantage of parallel environment, the application needs to be split into pieces. In this introductory chapter, you will learn about different ways of partitioning and how it fits different hardware configurations. You will also be introduced to various R packages that support parallel programming.

    Play Chapter Now
    Partitioning problems into independent pieces
    50 xp
    Partitioning demographic model
    50 xp
    Partitioning probabilistic demographic model
    50 xp
    Find the most frequent words in a text
    100 xp
    Models of parallel computing
    50 xp
    A simple embarrassingly parallel application
    100 xp
    Probabilistic projection of migration (setup)
    50 xp
    Probabilistic projection of migration
    100 xp
    R packages for parallel computing
    50 xp
    Passing arguments via clusterApply()
    100 xp
    Sum in parallel
    100 xp
    More tasks than workers
    100 xp
  2. 2

    The parallel Package

    This chapter will dive deeper into the parallel package. You'll learn about the various backends and their differences and get a deep understanding about the workhorse of the package, namely the clusterApply() function. Strategies for task segmentation including their pitfalls will also be discussed.

    Play Chapter Now
    Cluster basics
    50 xp
    Exploring the cluster object
    100 xp
    Socket vs. Fork
    100 xp
    The core of parallel
    50 xp
    Benchmarking setup
    100 xp
    Task size matters
    100 xp
    Initialization of nodes
    50 xp
    Loading package on nodes
    100 xp
    Setting global variables
    100 xp
    Exporting global objects
    100 xp
    Subsetting data
    50 xp
    Passing data as arguments
    100 xp
    Chunking migration application on worker's side
    100 xp
    Alternative chunking
    100 xp
  3. 3

    foreach, future.apply and Load Balancing

    In this chapter, you will look at two user-contributed packages, namely foreach and future.apply, which make parallel programming in R even easier. They are built on top of the parallel and future packages. In the last lesson of this chapter, you will learn about the advantages and pitfalls of load balancing and scheduling.

    Play Chapter Now
    foreach
    50 xp
    Combining results
    50 xp
    Word frequency with foreach
    100 xp
    Multiple iterators in word frequency
    100 xp
    foreach and parallel backends
    50 xp
    Using doParallel
    100 xp
    Word frequency with doParallel
    100 xp
    Word frequency with doFuture and benchmarking
    100 xp
    future and future.apply
    50 xp
    Word frequency with future.apply
    100 xp
    Planning future
    100 xp
    Benchmark future
    100 xp
    Load balancing and scheduling
    50 xp
    Load balancing
    100 xp
    Scheduling
    100 xp
  4. 4

    Random Numbers and Reproducibility

    Now you might ask, can I reproduce my results if the application uses random numbers? Can I generate the same results regardless of if the code runs sequentially or in parallel? This chapter will answer these questions. You will learn about a random number generator well suited to a parallel environment and how the various packages make use of it.

    Play Chapter Now
    Are my results reproducible?
    50 xp
    Reproducibility (I)
    50 xp
    SOCK vs. FORK
    100 xp
    SOCK vs. FORK & random numbers
    30 xp
    Parallel random number generators
    50 xp
    Setting an RNG
    100 xp
    Reproducible results in parallel
    100 xp
    Non-reproducible results in parallel
    100 xp
    Reproducibility in foreach and future.apply
    50 xp
    Reproducibility (II)
    50 xp
    Reproducing migration app with foreach
    100 xp
    Reproducing migration app with future.apply
    100 xp
    Next steps
    50 xp

Datasets

Words (Jane Austen's 6 books)US migrationSOTU (2016)

Collaborators

Benjamin Feder

Prerequisites

Writing Efficient R Code
Hana Sevcikova Headshot

Hana Sevcikova

Senior Research Scientist, University of Washington

Hana works as a senior research scientist at the University of Washington’s Center for Statistics and the Social Sciences, in the area of statistical computing. She also works as a data scientist at the Puget Sound Regional Council in Seattle. She has been involved in projects advancing parallel computing in statistics. She contributed to the snow package which became the R core package parallel. She developed a methodology for fault tolerant and reproducible parallel computing, implemented in snowFT, as well as the first interface to the L’Ecuyer's random number generator, the rlecuyer package. Currently, she’s been working in the area of statistical demography, developing R packages used by the United Nations for projecting demographic indicators for all countries of the world.
See More

What do other learners have to say?

    Join over 11 million learners and start Parallel Computing in R today!

    Create Your Free Account

    GoogleLinkedInFacebook

    or

    By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.