Revolutions

AI, Machine Learning and Data Science Roundup: November 2019

A 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 recently.

Open Source AI, ML & Data Science News

Python 3.8 is now available. From now on, new versions of Python will be released on a 12-month cycle, in October of each year.

Python takes the #2 spot in Github's annual ranking of programming language popularity, displacing Java and behind JavaScript.

PyTorch 1.3 is now available, with improved performance, deployment to mobile devices, "Captum" model interpretability tools, and Cloud TPU support.

The Gradient documents the growing dominance of PyTorch, particularly in research.

Keras Tuner, hyperparameter optimization for Keras, is now available on PyPI.

ONNX, the open exchange format for deep learning models, is now a Linux Foundation project.

AI Inclusive, a newly-formed worldwide organization to promote diversity in the AI community.

Industry News

Databricks announces the MLflow Model Registry, to share and collaborate on machine learning models with MLflow.

Flyte, Lyft's cloud-native machine learning and data processing platform, has been released as open source.

RStudio introduces Package Manager, a commercial RStudio extension to help organizations manage binary R packages on Linux systems.

Exploratory, a new commercial tool for data science and data exploration, built on R.

GCP releases Explainable AI, a new tool to help humans understand how a machine learning model reaches its conclusions.

Google proposes Model Cards, a standardized way of sharing information about ML models, based on this paper.

GCP AutoML Translation is now generally available, and the GCP Translation API is now available in Basic and Advanced editions.

GCP Cloud AutoML is now integrated with the Kaggle data science competition platform.

Amazon Rekognition adds Custom Labels, allowing users to train the image classification service to recognize new objects with as few as 10 training images per label.

Amazon Sagemaker can now use hundreds of free and paid machine learning models offered in Amazon Marketplace.

The AWS Step Functions Data Science SDK, for building machine learning workflows in Python running on AWS infrastructure, is now available.

Microsoft News

Azure Machine Learning service has released several major updates, including:

• a new web-based "studio" user interface (shown in this video),
• a new drag-and-drop workflow designer (video),
• a simple UI to apply automated machine learning to a dataset (video)
• built-in Notebook support,
• creation of end-to-end machine learning pipelines for training and deployment (video),
• auto-scaling compute instances for training and model deployment (video),
• a new Model Interpretability toolkit (video),
• a new asynchronous batch inferencing capability (video),
• integration with Azure Open Datasets (video),
• integration with Visual Studio Code,
• expanded PyTorch support, and
• new R SDK support (video).

Visual Studio Code adds several improvements for Python developers, including support for interacting with and editing Jupyter notebooks.

ONNX Runtime 1.0 is now generally available, for embedded inference of machine learning models in the open ONNX format.

Many new capabilities have been added to Cognitive Services, including:

• Personalizer, a new service to optimize user interfaces with reinforcement learning;
• Form Recognizer, a new service to extract structured data from printed or handwritten forms;
• A new Custom Neural Voice capability, to build a unique voice from a few minutes of training audio;
• Microsoft Video Indexer adds multi-language speech identification and transcription;
• Several services, including Face API, Custom Vision, and Translator Text are new available to Azure Free accounts

Bot Framework SDK v4 is now available, and a new Bot Framework Composer has been released on Github for visual editing of conversation flows.

SandDance, Microsoft's interactive visual exploration tool, is now available as open source.

Learning resources

An essay about the root causes of problems with diversity in NLP models: for example, "hers" not being recognized as a pronoun. 

Videos from the Artificial Intelligence and Machine Learning Path, a series of six application-oriented talks presented at Microsoft Ignite.

A guide to getting started with PyTorch, using Google Colab's Free GPU offer.

Public weather and climate datasets, provided by Google.

Applications

The Relightables: capture humans in a custom light stage, drop video into a 3-D scene with realistic lighting.

How Tesla builds and deploys its driving automation models with PyTorch (presentation at PyTorch DevCon).

OpenAI has released the full GPT-2 language generation model.

Spleeter, a pre-trained PyTorch model to separate a music track into vocal and instrument audio files.

Detectron2, a PyTorch reimplementation of Facebook's popular object-detection and image-segmentation library.

Find previous editions of the AI roundup here.

AzureR updates: AzureStor, AzureVM, AzureGraph, AzureContainers

Some major updates to AzureR packages this week! As well as last week's AzureRMR update, there are changes to AzureStor, AzureVM, AzureGraph and AzureContainers. All of these are live on CRAN.

AzureStor 3.0.0

There are substantial enhancements to multiple-file transfers (up and down). You can supply a vector of pathnames to storage_upload/download as the source and destination arguments. Alternatively if you specify a wildcard source, there is now the ability to recurse through subdirectories; the directory structure in the source will be reproduced at the destination.

There are revamped methods for getting storage properties and (user-defined) metadata for storage objects.

You can now use Azurite with AzureStor, by creating a service_specific endpoint (file_endpoint, blob_endpoint, adls_endpoint) with the Azurite URL. AzureStor will print a warning, but create the endpoint anyway.

For other changes, see the NEWS.md file.

AzureVM 2.1.0

You can now create VM scalesets with attached data disks. In addition, you can specify the disk type (Standard_LRS, StandardSSD_LRS, or Premium_LRS) for the OS disk and, for a Linux Data Science Virtual Machine, the supplied data disk. This enables using VM sizes that don't support Premium storage.

AzureGraph 1.1.0 and AzureContainers 1.1.2

These packages have been updated to use the new Microsoft Graph operations, introduced last week, for managing app passwords. As a security measure, app passwords can no longer be manually specified; instead they are auto-generated on the server using a cryptographically secure PRNG.

In AzureGraph, the az_app$update_password() method is defunct; in its place are add_password() and remove_password(). Similarly, in AzureContainers the aks$update_service_password() and aks$update_aad_password() methods no longer accept a manual password as an input.

If you use Graph, and in particular if you use AzureContainers to deploy Azure Kubernetes Service clusters, you should update as soon as possible, as the old versions are incompatible with the new Graph API.

If you run into any problems, please open an issue at the GitHub repo in question.

Azure AI and Machine Learning talk series

At last week's Microsoft Ignite conference in Orlando, our team delivered a series of 6 talks about AI and machine learning applications with Azure. The videos from each talk are linked below, and you can watch every talk from the conference online (no registration necessary). Each of our talks also comes with a companion Github repository, where you can find all of the code and scripts behind the demonstrations, so you can deploy and run them yourself.

If you'd like to see these talks live, they will also be presented in 31 cities around the world over the next six months, starting with Paris this week. Check the website for Microsoft Ignite the Tour for event dates and further information.

AIML10 Making sense of your unstructured data with AI

Tailwind Traders has a lot of legacy data that they’d like their developers to leverage in their apps – from various sources, both structured and unstructured, and including images, forms, and several others. In this session, learn how the team used Azure Cognitive Search to make sense of this data in a short amount of time and with amazing success. We discuss tons of AI concepts, like the ingest-enrich-explore pattern, search skillsets, cognitive skills, natural language processing, computer vision, and beyond.

Github repository here. Video here.

AIML20 Using pre-built AI to solve business challenges

As a data-driven company, Tailwind Traders understands the importance of using artificial intelligence to improve business processes and delight customers. Before investing in an AI team, their existing developers were able to demonstrate some quick wins using pre-built AI technologies.

In this session, we show how you can use Azure Cognitive Services to extract insights from retail data and go into the neural networks behind computer vision. Learn how it works and how to augment the pre-built AI with your own images for custom image recognition applications.

Github repository here. Video here.

AIML21 Developers guide to AI: A data story

In this theater session, we show the data science process and how to apply it. From exploration of datasets to deployment of services - all applied to an interesting data story. We also take you on a very brief tour of the Azure AI platform.

Github repository here. Video here.

AIML30: Start building machine learning models faster than you think

Tailwind Traders uses custom machine learning models to fix their inventory issues – without changing their software development life cycle! How? Azure Machine Learning Visual Interface.

In this session, learn the data science process that Tailwind Traders’ uses and get an introduction to Azure Machine Learning Visual Interface. See how to find, import, and prepare data, select a machine learning algorithm, train and test the model, and deploy a complete model to an API. Get the tips, best practices, and resources you and your development team need to continue your machine learning journey, build your first model, and more.

Github repository here. Video here.

AIML40 Taking models to the next level with Azure Machine Learning best practices

Tailwind Traders’ data science team uses natural language processing (NLP), and recently discovered how to fine tune and build a baseline models with Automated ML.

In this session, learn what Automated ML is and why it’s so powerful, then dive into how to improve upon baseline models using examples from the NLP best practices repository. We highlight Azure Machine Learning key features and how you can apply them to your organization, including: low priority compute instances, distributed training with auto scale, hyperparameter optimization, collaboration, logging, and deployment.

Github repository here. Video here.

AIML50 Machine learning operations: Applying DevOps to data science

Many companies have adopted DevOps practices to improve their software delivery, but these same techniques are rarely applied to machine learning projects. Collaboration between developers and data scientists can be limited and deploying models to production in a consistent, trustworthy way is often a pipe dream.

In this session, learn how Tailwind Traders applied DevOps practices to their machine learning projects using Azure DevOps and Azure Machine Learning Service. We show automated training, scoring, and storage of versioned models, wrap the models in Docker containers, and deploy them to Azure Container Instances or Azure Kubernetes Service. We even collect continuous feedback on model behavior so we know when to retrain.

Github repository here. Video here.

New R Support in Azure Machine Learning

Azure Machine Learning has added support for the R language, it was announced at the Ignite conference in Orlando this week.

A new R package azuremlsdk (available to install from Github now, and from CRAN soon), provides the interface to the Azure Machine Learning service. With R functions, you can provision new computing clusters in Azure, and use those to train models with R and deploy them as prediction endpoints for use from any app. You can also launch R-based notebooks in the new Azure Machine Learning studio web interface, or even launch a complete RStudio server instance on your cloud computing resources. Azure Machine Learning service supports the latest version of R (3.6.1) and all R packages (from CRAN, Github, or elsewhere). The video below from The AI Show demonstrates how it all works:

Azure Machine Learning is also great for teams that have both Python and R expertise. You can even call Python models from R (and vice-versa): in this Ignite 2019 talk (presented by me and Daniel Schneider) we deploy R and Python function as a container services, and call them both from a Shiny app. You can also find the slides and associated code from that talk in this Github repository.

To get started with R in Azure Machine Learning, a good place to start is the tutorial "Train and deploy your first model in R with Azure Machine Learning". If you need an Azure subscription, use this link to sign up for Azure and get $200 in free Azure credits.

Azure Machine Learning: ml.azure.com

AzureRMR 2.3.0 now on CRAN

This post is to announce that the latest update to AzureRMR is now available on CRAN. Version 2.3.0 brings several changes to make life easier when managing resources in Azure.

New in this version is a facility for parallelising connections to Azure, using a pool of background processes. Some operations, such as downloading many small files or interacting with a cluster of VMs, can be sped up significantly by carrying them out in parallel rather than sequentially. The code for this is currently duplicated in multiple packages including AzureStor and AzureVM; moving it into AzureRMR removes the duplication and also makes it available to other packages that may benefit. See the vignette for more details.

One side-effect of this change is that loading a current version of AzureStor or AzureVM, along with AzureRMR, will bring up a message in the R console:

> library(AzureRMR)
> library(AzureVM)
Attaching package: 'AzureVM'
The following objects are masked from 'package:AzureRMR':
    delete_pool, init_pool

Similarly, if you load the SAR package, you will receive a warning:

> library(SAR)
Warning messages:
1: replacing previous import 'AzureRMR::init_pool' by 'AzureStor::init_pool' when loading 'SAR'
2: replacing previous import 'AzureRMR::delete_pool' by 'AzureStor::delete_pool' when loading 'SAR'

These messages are because the pool functions in AzureRMR have the same names as those in the other packages. You can safely ignore them; everything will still function correctly, and I'll be submitting updated versions to CRAN in the next few days (as soon as the AzureRMR update propagates to CRAN mirrors).

Other changes in 2.3.0 include:

• Subscription and resource group objects now have do_operation methods, like resource objects. This allows you to carry out arbitrary operations on a sub or RG, if you know the REST call.
• AzureGraph is now a direct import, which should help ensure your credentials are consistent for Resource Manager and Microsoft Graph.
• Error messages should now be much more informative, especially when deploying templates.

If you run into problems, or to send feedback, please open an issue at the GitHub repo.