Microsoft launches machine learning as an Azure service

Microsoft's new Azure service aims to take the pain out of managing large machine learning systems

Microsoft has expanded the data-analysis offerings on its Azure cloud, offering a machine learning service to help organizations derive more insight from mountains of unstructured data.

The new service is designed to reduce the amount of work needed to deploy machine learning.

"Our focus has been simplicity, simplicity, simplicity," said T. K. "Ranga" Rengarajan, Microsoft's vice president of engineering for data platform, cloud and enterprise.

Machine-learning systems can be costly and time-consuming to set up and run in-house, due to the complexities of the software and the large amount of hardware required for the task. The cloud provides an easy way for enterprises to ramp up large-scale machine learning jobs, Rengarajan said.

Microsoft first launched Machine Learning as a preview in June 2014, though did not state when the service would go live. Early users include the Pier 1 retail chain, Carnegie Mellon University, and energy industry systems provider eSmart Systems.

Another early Microsoft Machine Learning user has been Ziosk, which provides restaurants like Chili's with tableside tablets that customers can use to order menu items. Ziosk used machine learning in conjunction with the Microsoft Power BI service to provide restaurants with more information about customer purchasing habits and trends.

Machine learning is a type of data analysis that allows computers to draw inferences through large sets of data. While IBM has incorporated machine learning into some of its Watson-branded cognitive computing services, Microsoft's offering is the first general machine learning service from a major cloud vendor.

With the Azure service, programmers can use either the R programming language or Python. Machine learning algorithms may be purchased in the Microsoft Azure Marketplace, or obtained for no cost in a new community gallery Microsoft has created for users to share their formulas.

Machine Learning, which will be available as a full service April 1, will cost $9.99 per seat per month, $1 per hour for use of the developer environment, $2 per every compute hour, and $0.50 per every 1,000 transactions.

Microsoft has also unveiled a number of other Azure data analysis services.

Azure now offers the ability to run Hadoop over Ubuntu Linux. This feature provides a way for Hadoop deployments that use Linux scripts to run on Azure.

Azure also offers a hosted version of Storm, open source software for analyzing data streams. Azure allows developers to easily connect .Net and Java libraries to Storm.

Online ad provider Linkury has used the Storm service to track customer interactions, according to Microsoft.

From CIO: 8 Free Online Courses to Grow Your Tech Skills
View Comments
Join the discussion
Be the first to comment on this article. Our Commenting Policies