Machine Learning announcement

• Link
  Azure Machine Learning now supports Azure DocumentDB as a data source in Import Data module
  GokhanU Monday, November 28, 2016 6:28 PM

  We have released support for Azure Document DB as a data source in Azure Machine Learning. You can use the existing "Azure DocumentDB" connection option in the Import Data module to read data from Azure DocumentDB for your experiment.
  For more information, please see the DocumentDB section of the Import Data module.
• Link
  Updates to Text Analytics Modules in Azure Machine Learning Studio
  Roope Astala - MSFT Thursday, October 20, 2016 1:39 PM

  New Module: Extract Key Phrases from Text

  You can use this module to extract key talking points from text. As an input, the module takes a dataset that must have a text string column from which the key-phrases are extracted.

  The module takes the language of the text records as input parameter. Supported languages include Dutch, English, French, German, Italian and Spanish. You can also use a language column that specifies the language of each record, as produced by Detect Languages module.  

  The output consists of comma-separated lists of key phrases for each record in input. The key phrases can be used to summarize a corpus of documents, or as features for a machine learning model.

  Updated Module: Preprocess Text

  • You can specify a language through a language column, as produced by Detect Languages module.
  • Following three preprocessing options have been added: Expand verb contractions, Normalize backslashes to slashes, and Split tokens on special characters. Previously, these transformations were done automatically.
• Link
  Azure Machine Learning Workspace and Web Service Pricing Plans now available in the Azure Portal
  Chhavi Bhasin Wednesday, August 24, 2016 10:02 PM

  We are pleased to announce the availability of Azure Machine Learning Workspaces and Web Service Plans for all our Azure Machine Learning users through the Azure Portal. Azure Machine Learning users can now create and manage Standard workspaces through the Azure Portal. In addition, users will also be able to create Web Service Pricing Plans. These plans are used when deploying web services and provide included quantities of operationalized compute at a single, predictable monthly cost.

  Create your Standard Azure Machine Learning workspace now by going to https://portal.azure.com. Log in with the credentials that you use for accessing your Azure Subscription(s). Click on +New | Data + Analytics | Machine Learning Workspace.

• Link
  New Text Analytics Modules in Azure ML Studio
  Roope Astala - MSFT Thursday, August 11, 2016 7:47 PM

  We are pleased to announce significant new capabilities for text analytics in Azure Machine Learning Studio.

  The new features include following modules:

  • Detect Languages
    • Identify language of each record in input file from large number of languages.
  • Preprocess Text
    • Clean and simplify text to make it more easy to featurize.
  • Extract N-Gram Features from Text
    • Create N-gram feature vectors from long text strings, and select only the most important features.
  • Latent Dirichlet Allocation
    • Group text into categories using topic modeling.

  These modules allow you to build models to solve text classification problems, such as support ticket routing or sentiment analysis. You can pre-process text in multiple languages, and then create features from your text data. Operationalization of models is fully supported.

  The modules complement the existing capabilities for Feature Hashing, Vowpal Wabbit based high-dimensional models, and text analytics through R and Python scripting.

  For more details, visit MSDN documentation and Cortana Intelligence Gallery.

• Link
  Web services created using the "New" option
  raymondl_msft Friday, July 22, 2016 7:59 PM

  There is an issue impacting the "New" web service option for deploying web services from Predictive Experiments in Azure ML. We are working on resolving the issue, and a result have disabled the feature until the feature is fully functional. To access web services created the new process, please browse to https://services.azureml.net and sign in to view your web services. Sorry for any inconvenience this issue may cause.

• Link
  Azure Machine Learning now supports Azure SQL Data Warehouse as a data source and a destination
  GokhanU Thursday, March 24, 2016 2:24 PM

  We have released support for Azure SQL Data Warehouse as a data source and a destination in Azure Machine Learning. You can use the existing "Azure SQL Database" connection options in the Reader and Writer modules to read from and write to Azure SQL Data Warehouse. When using the Writer module, the destination tables must already exist in the SQL Data Warehouse.

  For more information, please see How to Use Azure ML with Azure SQL Data Warehouse 

  Please refer to SQL Data Warehouse Reference to learn more about the product and the Transact-SQL language details.

   

• Link
  Announcing Availablility of Tree Model Visualizations in Azure Machine Learning
  Roope Astala - MSFT Monday, November 23, 2015 9:32 PM

  Visualization of tree models such as Boosted Decision Trees is now available in Azure Machine Learning Studio. To view the trees, train the model, and click Visualize on the output of Train Model module.
• Link
  Announcing the Availability of an Azure Virtual Machine Image with Popular Data Science Tools
  Larry Koch (MSFT) Thursday, November 19, 2015 12:59 PM

  Announcing the Availability of an Azure Virtual Machine Image with Popular Data Science Tools

  Microsoft Data Group is happy to announce the immediate availability of a Windows Server 2012 based custom virtual machine image on the Azure marketplace containing several tools that can be used by data scientists and developers for advanced analytics. Through Azure’s world-wide cloud infrastructure, customers now have on-demand access to a data science development environment they can use to derive insights from their data, build predictive models and intelligent applications.  The virtual machine saves developers’ time from having to discover and install the tools individually.  Hosting the data science machine on Azure gains you high availability and a consistent set of tools used across your data science team.

  The data science VM comes with several popular tools pre-installed like Revolution R Open, Anaconda Python distribution including Jupyter notebook server, Visual Studio Community Edition, Power BI Desktop, SQL Server Express edition and Azure SDK. Once you provision your virtual machine from this image you can get started with data exploration and modeling right away. The data on the virtual machine is stored on the cloud and highly available. You have full administrative access to the virtual machine and can install additional software as needed. There is no separate software fee to use the VM image. You only pay for actual hardware compute usage of the virtual machine depending on the size of the virtual machine you are provisioning this VM on. You  can turn off the machine from Azure portal when it is not in use to avoid being billed. When you restart the virtual machine from the Azure portal you can continue your development with all your data and files intact. Further augment your analytics on your data science virtual machine by leveraging solutions in Microsoft’s Cortana Analytics Suite.

  The data science virtual machine helps you create an analytics environment where you can rapidly build advanced analytics solutions for deployment to the cloud, on-premises or in a hybrid environment.

  You can find the data science virtual machine and the Azure hardware compute pricing  at: https://azure.microsoft.com/en-us/marketplace/partners/microsoft-ads/standard-data-science-vm/

  More information about the virtual machine can be found at: https://azure.microsoft.com/en-us/documentation/articles/machine-learning-data-science-provision-vm/

  If you are new to Azure, you can try the data science virtual machine for free via a 30-day Azure free trial by visitinghttps://azure.microsoft.com/en-us/pricing/free-trial/

  We encourage you to try the data science virtual machine to jumpstart your analytics project and provide us feedback on how we can better serve your analytics needs.

• Link
  Azure Machine Learning now available in Western Europe
  Dan Manrique Wednesday, September 09, 2015 9:10 PM

  We are happy to announce that we have released Azure ML in our Western Europe datacenter (Amsterdam). Now you can create workspaces in this datacenter. For more information, click here: http://aka.ms/mlwelaunch.

• Link
  Azure Machine Learning now available in Southeast Asia
  Dan Manrique Wednesday, September 09, 2015 9:09 PM

  We are happy to announce that we have released Azure ML in our SouthEast Asia datacenter (Singapore). Now you can create workspaces in this datacenter. For more information, click here: http://aka.ms/mlasialaunch.
• Link
  Azure Machine Learning now supports Azure Active Directory
  Dan Manrique Wednesday, September 09, 2015 9:09 PM

  We are happy to announce that we have released Azure Active Directory (AAD) support in Azure ML. Now you can log in with any arbitrary Azure AD account (work or school account), in addition to, Microsoft accounts (LiveID), and invite other Azure AD users to your workspace. For more information, click here: http://blogs.technet.com/b/machinelearning/archive/2015/09/02/logging-on-to-azure-ml-with-your-work-or-school-account.aspx.

• Link
  Free Excel add-in to connect to Azure ML web services
  Ted Way Wednesday, September 02, 2015 6:52 PM

  A free Excel add-in that you can use with web services published from Azure Machine Learning is now available. You can use this add-in for request/response predictions or batch predictions, work in Windows or the browser, share workbooks with your co-workers, and call multiple web services all within a single spreadsheet.  Go to http://aka.ms/amlexcelhelp for help or ask a question here.

  To try it out, open and download sample Excel worksheets that already contain web services:

  http://aka.ms/amlexcel-sample-1

  http://aka.ms/amlexcel-sample-2

  You may use the add-in directly in the browser using Excel Online or opening the file in Excel 2013 or later on Windows.  Copy the file to your own OneDrive account if you want to edit it.

  Feature highlights

  • Connect to multiple web services in one Excel workbook
  • Choose from RRS or BES
  • Supports single or no input, and single, multiple, or no outputs

  For sample 1 (text sentiment analysis): http://aka.ms/amlexcel-sample-1

  1.)    Highlight cells A1 to A12

  2.)    Click the range selector button (the selection Sheet1!$A$1:$A:$12 should automatically be populated)

  3.)    Click OK in the Select Data dialog box

  4.)    Type “B1” in the output1 text box

  5.)    Click the Predict button

  6.)    This web service takes some time to process the text, so please be patient and wait for a minute.  When it’s done, you should see the sentiment predictions and scores in columns B and C.

  For sample 2 (Titanic survivor predictor): http://aka.ms/amlexcel-sample-2

  1.)    Highlight cells A1 to G11

  2.)    Click the range selector button (the selection Sheet1!$A$1:$G:$11 should automatically be populated)

  3.)    Click OK in the Select Data dialog box

  4.)    Type “H1” in the output1 text box

  5.)    Click the Predict button

  6.)    When it’s done, you should see the predictions and scores in columns H and I

  To add your own web service:

  1.)    In the Excel add-in, go to the Web Services section (if you are in the Predict section, click the back arrow to go to the list of web services)

  2.)    Click Add Web Service

  3.)    In Azure ML Studio, click the WEB SERVICES section in the left pane, and then select the web service

  4.)    Copy the API key for the web service

  5.)    Paste the API key into the Excel add-in text box labeled API key

  6.)    On the DASHBOARD tab for the web service, click the REQUEST/RESPONSE link

  7.)    Look for the OData Endpoint Address section. Copy the URL and paste that into the text box labeled URL in the Excel add-in\

  8.)    Click Add

• Link
  Release Announcement: Preview availability of Jupyter Notebooks in Azure Machine Learning Studio
  Dan Manrique Friday, July 24, 2015 4:19 PM

  On July 24th, 2015, Microsoft announced the Preview Availability release of Jupyter Notebooks in Azure Machine Learning Studio. 
   
  Azure Machine Learning Studio is a powerful canvas for the composition of Machine Learning Experiments and subsequent operationalization and consumption.   It provides an easy to use, yet powerful, drag-drop style of creating Experiments.  But sometimes you need a good old “REPL” that allows you to have a tight loop where you enter some script code and get a response.  We are delighted to announce that we’ve now integrated this functionality into ML Studio through Jupyter Notebooks.
   
  Jupyter enables the concept of “executable documents” with support for mixed code, markdown and inline graphics.   It’s  one of the most important innovations in the Data Science and Technical Computing space in recent years.  You now have full access to its power from any OS, from any modern browser directly from inside the Azure Machine Learning Studio. 
   
  In addition to authoring capabilities above, we are also enabling publishing AzureML web services directly from the Jupyter Notebook. We are also extending this capability to the Jupyter Notebooks running locally outside of AzureML Studio. This allows you to publish any function, including those creating ML models, to be published as a web service directly from the Jupyter Notebook running on your machine. The result is an AzureML web service API that can be called to perform functions or predictions from client applications in real time and over the internet.
• Link
  Annoucing the availability of Azure Machine Learning BES SDK
  raymondl_msft Monday, June 29, 2015 5:24 PM

  Announcing the availability of the SDK for AzureML Batch Execution Service (BES)

  The AzureML BES SDK is now available for download and installation as a NuGet package on NuGet.org (http://www.nuget.org/packages/Microsoft.Azure.MachineLearning/).

  The SDK wraps the BES sample code with additional functions to simplify the consumption of BES APIs.

  Documentation is available after installing the SDK package in Visual Studio. The BES documentation has also been updated with sample code and guidance on using the SDK.

  We are looking forward to hearing your feedback and comments on the SDK to help improve it.

  Thanks,

  AzureML Team

  
  
• Link
  AzureML Retraining API demo available on Codeplex
  raymondl_msft Tuesday, March 24, 2015 6:18 PM

  We have posted a demo of the Retraining APIs on Codeplex.com. The demo uses the new APIs to programmatically retrain a trained model. Here is the link to the Demo.

  Please take a look and let us know if you have questions or comments.