Machine Learning: Azure Machine Learning


(This post is translated from Olivia’s Machine Learning Series)

In this blog post I am going to give a brief overview of Azure Machine Learning (Azure ML) and what is the vision behind it.


At the Strata + Hadoop World in San Jose on February 18, Joseph Sirosh, CVP machine learning at Microsoft, announced the general availability of the Azure ML Service: A cloud service that allows you to quickly and easily create predictive analytics models based on existing machine learning algorithms. The vision of Azure ML is to democratize, to provide access to data and machine learning to everyone in the enterprise – analysts, developers, departments, etc.



The typical workflow of Azure ML begins with the import of the data. Both sources are possible: both the cloud and locally stored files. In the cloud, there are several connecting points such as a SQL Server VM, Blobs or Tables, an Azure SQL database, HDInsight-Cluster, Web URL (via HTTP) or a Data Feed Provider. Local data can also be found by uploading them to Azure. More information about the data ingress will be given later in a separate blog post.

Machine Learning Studio: Cleaning up data, Building a model

To finally start with the given data and based on that to choose which data models to build, there is the ML Studio.


The ML Studio is a browser-based application: No software to install, no hardware to buy, no VM to start; only the browser of your choice.

In ML Studio there is the possibility to set up different workspaces, which are equivalent to sandboxes:


Why? One reason is that you can nicely distinguish its models according to business reason. On the other, you can share your workspaces with colleagues to work on data models together:

You can also invite other people as “owners”. The big difference between an “owner” and a “user” is that you as an owner can invite more people to the workspace.


When modeling yourself, you can access a catalogue of proven ML algorithms. With drag & drop you can easily assemble a model from a catalogue of modules; almost like Visio for data scientists. The selection of existing machine learning algorithms has resulted from decades of research in Microsoft Research, Halo, Kinect and Bing.


Those who have already created their models in proven data science open source programming languages, such as R and Python, do not have to start from the beginning at Azure ML. On the contrary, R and Python are supported! Using R or Python modules, you can add your previously created R-/Python-Code to Azure ML, according to the motto Microsoft < 3 Open Source.



What has so far been weeks, even months before a model has been integrated into the production code, can be accelerated at Azure ML by providing a model as a Web service in a matter of minutes:


Such a Web service can then be built into dashboards, apps, or other applications through API authentication. Example code in C, R and Python is provided.

How do I start with Azure ML?

You can try Azure ML without a free Azure subscription with a Microsoft account (i.e. Outlook, Hotmail, Yahoo or live email address). Just click on the and get started now. More information can be found here.

If you have a Azure subscription, you will find a grid for machine learning in the Azure Management Portal.

A storage account must then be specified or created to store the uploaded local data.
Once a workspace is created, the “Open in Studio” button will take you to the ML Studio.

Anh Vo

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s