Azure Machine Learning Aims to Convert Data to Information

Microsoft on Wednesday announced new data services running on its Azure cloud in what it has positioned as a bid to bring big data to the mainstream.

Those services include the HDInsight Apache Hadoop-based service; Storm on HDInsight, which lets users use Hadoop and Storm to create distributed, real-time data processing solutions in Azure; and Azure Machine Learning, a managed cloud service for advanced analytics.
Azure Machine Learning lets users build and deploy apps and conduct predictive analyses, among other things.

Those tasks are facilitated by the Machine Learning Marketplace, or MLM, which offers analytics services, algorithms and APIs users can plug into their solutions.
Microsoft also has signed up several partners for Azure, such as Informatica and Ziosk.
"The Azure solution is much more intuitive than existing network and platform management solutions, especially for nontechnical people," noted Jim McGregor, principal analyst at Tirias Research.

 

Microsoft and Machine Learning

Azure Machine Learning can be set up using a Web browser. Users make drag-and-drop gestures and create simple flow graphs to set up experiments.
Users can pull sample experiments, packages written in R and Python, and best-in-class algorithms from Xbox and Bing -- or they can write their own custom code in R or Python.
Algorithms such as "Learning with Counts" let users learn from terabytes of data on Microsoft's servers.
Users can easily update the models they build and return them to production. They can share their solutions with the Azure Machine Learning community in the product gallery, or monetize and brand them for sale in the Azure MLM.

Big Data, Machine Learning and Microsoft

Machine learning comes from intelligent recognition of patterns by complex algorithms. Those algorithms have to work on large datasets, so big data "is critical" to machine learning, McGregor told TechNewsWorld.
However, Microsoft's machine learning "seems more like object-oriented design than machine learning," he pointed out.
"While there is no set definition as to what level of intelligence constitutes machine learning, the Azure solution appears to stretch the term beyond what most would consider true machine-level intelligence," McGregor said, because the algorithms appear to really be networking models.
Microsoft "alludes to learning through data analytics, but doesn't provide any details," McGregor remarked.
Nevertheless, Microsoft's tools let users "quickly and easily define and configure cloud-based resources without the need of dedicated IT personnel," he pointed out. Viewed that way, Microsoft offers "a very interesting and useful platform."
However, those tools don't eliminate reliance on IT staff, because IT professionals would be needed to configure solutions effectively, McGregor cautioned, especially when custom code is required.

Transforming Data Into Information

The next wave of IT is going to be information technology, where people will work interactively with their devices, predicted Mike Jude, a program director at the Stratecast service of Frost & Sullivan. Services such as those Microsoft introduced play into that.
"What we've had so far is the data age," Jude told TechNewsWorld. "We've learned how to generate and store data -- but so far, automation hasn't been able to turn that into information, which is data in context. That's why you need cognitive computing, machine learning and heuristics, and stuff like that."

Playing Catch-Up With IBM

IBM has been offering big data solutions for small businesses for a while, and that's "sort of like what Microsoft's trying to market," Jude said.
IBM last year announced Watson Analytics, a natural language-based cognitive service that makes advanced and predictive analytics easy to acquire and use. A freemium version was released for the desktop and mobile devices.