Revolutions

Because it's Friday: Happy Holidays

🎵 He'd better watch out // he'd better comply 🎵:

This is my favourite festive GDPR gag of Christmas 2018 so far. pic.twitter.com/2HheWIkviV

— Charlie King (@charlietheking) November 7, 2018

With the holiday season upon us, we'll be taking a little break here at the blog to relax and spend time with the family. We hope you get to do the same, and best wishes to all our readers for the season. We'll be back early in the new year: Happy New Year to all!

R 3.5.2 now available

R 3.5.2, the latest version of the R language for statistical computation and graphics from the R Foundation, was released today. (This release is codenamed "Eggshell Igloo", likely in reference to this or this Peanuts cartoon.) Compared to R 3.5.1, this update includes only bug fixes, so R scripts and packages compatible with R 3.5.0 or R 3.5.1 should work without modification.

Installers for R on Windows and Debian Linux are available now; other platforms should be available from CRAN within the next few days. If you want to see the full list of fixes, check out the announcement from the R Core Group at the link below.

R-announce mailing list: R 3.5.2 is released

AI, Machine Learning and Data Science Roundup: December 2018

A monthly roundup of news about Artificial Intelligence, Machine Learning and Data Science. This is an eclectic collection of interesting blog posts, software announcements and data applications from Microsoft and elsewhere that I've noted over the past month or so.

Open Source AI, ML & Data Science News

PyTorch 1.0 has been released with faster performance, better support for distributed computing, a new C++ frontend, and a model-sharing repository.

Tensorflow 2.0 will include a built-in implementation of Keras for deeper integration and support of eager execution.

Industry News

The AI Now Institute has published the AI Now Report 2018, and makes 10 recommendations related to the social impact of AI technology.

Stanford's AI Index 2018 Annual Report documents the year's progress in AI teaching, research, and humanlike capabilities.

Tableau adds SSL support for connecting the R server for its R integration feature.

AWS Lambda adds support for Python 3.7, and offers a customization framework for supporting other programming languages.

New features in Amazon SageMaker: semantic segmentation, early stopping for automatic model tuning, and tools for tracking machine learning experiments.

H20.ai's machine learning platform now available as a hosted service on AWS.

Microsoft News

Microsoft calls for government regulation on the use of facial recognition technology.

The Microsoft Connect event on December 4 included several product announcements related to Machine Learning and AI, summarized below.

Azure Machine Learning service is now generally available. This service orchestrates the training and deployment of machine learning modules, using resources managed on-premises or in Azure. Capabilities include:

• Model training with open-source frameworks including scikit-learn, PyTorch and Tensorflow.
• A Python module for managing and orchestrating models and compute resources.
• Integration with developer tools including Jupyter Notebooks and Visual Studio Code.
• The ability to launch and redirect training to CPU and GPU-enabled resources: local, Azure virtual machines, and distributed clusters with auto-scaling capabilities.
• An AI-driven service for testing and optimizing multiple ML algorithms, and hyperparameter tuning for PyTorch and Tensorflow models.
• Production deployment for trained models via containers and the open ONNX standard.
• Pipelines to orchestrate the entire machine learning process from data to deployment.

Azure Notebooks, the hosted Jupyter Notebooks service, has been updated. New features include:

• A refreshed user interface, with improved project management and notifications.
• Upgrades to the free compute image, and the ability to use paid virtual machine (including GPU-class) instances to provide the compute kernel.
• Streamlined integration with Azure Authentication and Azure Machine Learning service.

ONNX Runtime, a cross-platform, high-performance engine for inferencing with trained ML models in the Open Neural Network Exchange (ONNX) representation, has been released as open source.

ML.NET 0.8, the open-source machine learning framework for .NET, is now available with new recommendation scenarios and a feature importance tool.

The Language Understanding service can now be deployed in a self-hosted container.

The Translator Text API now offers customized translation to support specialized expressions and vocabulary.

Neural Text to Speech, a preview feature of Cognitive Services Speech Service, synthesizes spoken voices indistinguishable from real people.

Azure Functions now supports serverless deployment of Python functions. This blog post provides detailed examples.

Azure SQL Database now provides integration with the R language (in preview).

AzureR: a new suite of packages to manage Azure services with R.

Learning resources

The book Bayesian Methods for Hackers is now available as a series of Jupyter Notebooks with code in Tensorflow.

A tutorial on causal inference in computing systems, presented by Amit Sharma and Emre Kicima at KDD 2018.

A guide to resources and tutorials for getting started with machine learning on Azure.

A tutorial on packaging Cognitive Services APIs into containers and running them on a Docker-enabled platform.

Pattern Recognition and Machine Learning, the 2006 book by Christopher M Bishop, is now available for free download.

A 4-part series on the history and recent advancements in neural word embeddings.

Applications

Using a computer vision model to measure building footprints in satellite images.

Using active learning to track the migration of endangered bird species.

Using deep learning to improve the effectiveness of diabetic retinopathy models.

Using computer vision to quantify fish stocks in Australia.

A collection of reference architectures for deploying machine learning and AI models to real-time applications and batch processes with Azure.

Find previous editions of the monthly AI roundup here.

AzureStor: an R package for working with Azure storage

by Hong Ooi, senior data scientist, Microsoft Azure

A few weeks ago, I introduced the AzureR family of packages for working with Azure in R. Since then, I’ve also written articles on how to use AzureRMR to interact with Azure Resource Manager, how to use AzureVM to manage virtual machines, and how to use AzureContainers to deploy R functions with Azure Kubernetes Service. This article is the next in the series, and covers AzureStor: an interface to Azure storage.

The Resource Manager interface: creating and deleting storage accounts

AzureStor implements an interface to Azure Resource Manager, which you can use manage storage accounts: creating them, retrieving them, deleting them, and so forth. This is done via the appropriate methods of the az_resource_group class. For example, the following code shows how you might create a new storage account from scratch.

library(AzureStor)

# get the resource group for the storage account
rg <- AzureRMR::az_rm$
    new(tenant="{tenant_id}", app="{app_id}", password="{password}")$
    get_subscription("{subscription_id}")$
    get_resource_group("myresourcegroup")

# create the storage account
# by default, this will be in the resource group's region
rg$create_storage_account("mynewstorage")

Without any options, this will create a storage account with the following parameters:

• General purpose account (all storage types supported)
• Locally redundant storage (LRS) replication
• Hot access tier (for blob storage)
• HTTPS connection required for access

You can change these by setting the arguments to create_storage_account(). For example, to create an account with geo-redundant storage replication and the default blob access tier set to “cool”:

Because it's Friday: CGI you never knew was CGI

Computer-generated imagery in movies has gotten so good these days, much of the time you don't even realize it's there. You probably never noticed how Michael Cera's physique had been altered, or how Lost in Translation used motion capture technology from the future.

That's all from the blog team for this week. Have a great weekend, and see you next week!

In case you missed it: November 2018 roundup

In case you missed them, here are some articles from November of particular interest to R users.

David Gerard assesses the plausibility of a key plot point in 'Jurassic Park' with simulations in R.

In-database R is available in Azure SQL Database for private preview. 

Introducing AzureR, a new suite of R packages for managing Azure resources in R.

The AzureRMR package provides an interface for Resource Manager.

Roundup of AI, Machine Learning and Data Science news from November 2018.

You can now use the AI capabilities of Microsoft Cognitive Services within a container you host.

A look back at some of the R applications presented at the EARL conference in Seattle.

Slides and notebooks from my ODSC workshop, AI for Good.

T-Mobile uses AI models implemented with R to streamline customer service.

A guide to R packages for importing and working with US Census data.

Azure Machine Learning Studio, the online drag-and-drop data analysis tool, upgrades its R support.

And some general interest stories (not necessarily related to R):

• How the planets would look in the sky, if they were as close as the Moon.
• Pavarotti and Freddie Mercury, impersonated by the same performer.
• The real-world physics of the sci-fi series The Expanse
• Real robots and simulated people, dancing
• The different ways human languages write animal sounds

As always, thanks for the comments and please send any suggestions to me at [email protected]. Don't forget you can follow the blog using an RSS reader, via email using blogtrottr, or by following me on Twitter (I'm @revodavid). You can find roundups of previous months here. 

How to deploy a predictive service to Kubernetes with R and the AzureContainers package

by Hong Ooi, senior data scientist, Microsoft Azure

It's easy to create a function in R, but what if you want to call that function from a different application, with the scale to support a large number of simultaneous requests? This article shows how you can deploy an R fitted model as a Plumber web service in Kubernetes, using Azure Container Registry (ACR) and Azure Kubernetes Service (AKS). We use the AzureContainers package to create the necessary resources and deploy the service.

Fit the model

We’ll fit a simple model for illustrative purposes, using the Boston housing dataset (which ships with R in the MASS package). To make the deployment process more interesting, the model we fit will be a random forest, using the randomForest package. This is not part of R, so we’ll have to install it from CRAN.

data(Boston, package="MASS")
install.packages("randomForest")
library(randomForest)

# train a model for median house price
bos_rf <- randomForest(medv ~ ., data=Boston, ntree=100)

# save the model
saveRDS(bos_rf, "bos_rf.rds")

Scoring script for plumber

Now that we have the model, we also need a script to obtain predicted values from it given a set of inputs:

How to give money to the R project

by Mark Niemann-Ross, an author, educator, and writer who teaches about R and Raspberry Pi at LinkedIn Learning

I spend a LOT of time at r-project.org, in particular the sections for documentation and CRAN. But I hadn’t spent much time in the other areas: R Project, R Foundation, and links. When I recently wandered into the foundation area, I discovered something worth talking about: donations to the R Foundation.

I benefit from the work of the R Foundation. They oversee the language, but also encourage a healthy ecosystem. CRAN happens because of them. Updates to R happen because of them. useR! happens because of them. Every day, you and I are the recipients of some part of their time.

The least we can do is show them some appreciation. If you point your web browser at https://www.r-project.org/foundation/donations.html you’ll find a convenient (and surprisingly inexpensive) place to express your appreciation. As an individual, you can send these kind folks twenty-five euros to tell them you’re in favor of what they do.

You can send them more if you like. You can make your twenty-five euros an annual event. If you’re an institution, you can give 250 euros. If you’re really suave and sophisticated, you can give 500 euros. I suspect you could give more, but you may have to warn the R Board of Directors so they aren’t wondering what you’re up to. (kidding!)

Best of all – for your minimum of twenty-five euros you get listed on the R Donors Site. Look close – you’ll even see my name listed amongst the R glitterati.

Tax Deductible?

For those of us in the United States, donations to charitable organizations are tax deductible. However, The R Foundation is not recognized as such in the US. But take heart — The Society of Indian Academics in America has stepped in to provide a method for earmarking donations — and making donations deductible. Look at the very bottom of the foundations donor page for details on how to make this happen.

Take Action

Calculate the difference in time spent between working with R — and working without R — then divide twenty-five euros by that amount of time. If you’re like me, that becomes a pretty good value. I think it’s well worth the time to make a donation and keep the R Foundation feeling appreciated.

R Project: Membership Fees and Donations

Reflections on the 10th anniversary of the Revolutions blog

On December 9 2008, very nearly ten years ago, the first post on Revolutions was published. Way back then, this blog was part of a young startup called Revolution Computing, which later became Revolution Analytics. (That name persists to this day in the URL of this blog.) The idea at that time was to introduce the world to a wonderful but little-known statistics environment called R, by sharing news and applications of R to the broader data analysis community, along with tips and tricks for those that had already discovered R.

Since then, this blog has seen several waves shaking up the industry, from big data to data science, predictive analytics and machine learning, cloud computing and -- lately -- artificial intelligence. It's seen rising stars become leaders, new players become standards, and industry stalwarts enter the fray. And R has been there through it all.

I'd like to thank the organizers of the amazing SatRDays DC conference last weekend for allowing me to take some time at that event to reflect on the past 10 years of this blog, the R community, and how they intersected. You can see the video of the talk here. I've also embedded the slides below: they don't mean much without the talk track, but you can use them to appreciate Ashely McNamara's gorgeous art.

Oh, and here are links to the posts mentioned:

• The #3 post, "Using R with Jupyter Notebooks"
• The #2 post, "10 tips for making your R graphics look their best"
• The all-time most viewed post, "How to choose a random number in R"
• The Because it's Friday series (including that Bonnie Tyler post)

It's been an amazing 10 years. I'd like to thank my various managers over the years for supporting me in the blog, all of my guest bloggers and most of the commenters (but not the trolls), and especially you, dear readers, for making this all worthwhile. Blogging has been a big part of my life, and I look forward to being able to continue the tradition in the future as well.

Happy 10th birthday, Revolutions!

SatRDays DC: Afternoon Sessions videos (the Revolutions talk starts at 2:19:00)

Because it's Saturday: Go Your Own Way

I was delivering a workshop for AI Live yesterday so I didn't get the chance to do my Friday post, but I'm here at SatRDays DC and the playlist on the audio while we're waiting for things to start is amazing. Fleetwood Mac's Go Your Own Way just came on, which reminded me of this fascinating episode of Song Exploder where Lindsey Buckingham relates the story behind the song and how the musical components were layered together to crate the 1976 classic. 

We'll be back next week -- have a great weekend!