Behind the scenes at Labs
cancel
Showing results for 
Search instead for 
Did you mean: 

Discover 2016: Open the toolbox for cognitive computing

Curt_Hopkins

toolbox.gif

By Curt Hopkins, Managing Editor, Hewlett Packard Labs

Today, we’re excited to announce that we’ve released our Cognitive Computing Toolkit on GitHub. This release is the culmination of a long investigation into alternative forms of computing and will, in turn, become the foundation stone of future technology.

The imminent end of Moore’s Law – the doubling of transistors on a chip every 18 months to two years – looms large in the mind of researchers at Hewlett Packard Labs. According to Stan Williams, HPE Senior Fellow and director in the Foundational Technologies lab, one of the ways that we can continue to increase processing despite the collapse of Moore’s Law is “brain-like computing.”

“If we can adapt an understanding of how the brain works to our computing systems we can make improvements in the kinds of computers you and I care about,” Williams explained.

This approach led to the development of the platform code-named CogX. CogX depends on cognitive computing, rather than being strictly neuromorphic, though the ideas are related.

Cognitive vs. neuromorphic computing

“Cognitive’ generally implies biologically-inspired software or algorithms,” explained Ben Chandler, researcher in Software and Analytics. “CogX started life as a simulator and compiler for a brain-like chip but as we've moved farther from custom brain-like hardware, the project has become less neuromorphic and more cognitive.”

Kirk Bresniker, Labs’ Chief Architect for The Machine, described a cognitive model as “a system to model the continuous variables and the complex interactions via floating point math and differential field equations. Neuromorphic computing like the Memristor-based spiky neuron models directly represents the neurologic functions.”

 In April, HPE announced it was going to release an open-source version of that platform called the Cognitive Computing Toolkit and today the code for the tools was released on the code-sharing site GitHub.

Open the tool box

The Cognitive Computing Toolkit is a powerful open source software framework for building massively-parallel applications. Instead of relying on the traditional CPUs that power most computers, it does this running on graphics processing units (GPUs), inexpensive chips designed for video game applications.

Rich Friedrich, director of systems software for The Machine at Hewlett Packard Labs, gave a detailed explanation of the Cognitive Computing Toolkit at Discover London last year.

With the Cognitive Computing Toolkit, developers will have access to an easy-to-use, open-source platform designed for machine learning and data analysis applications. It will allow companies to leverage extreme performance, scale, and flexibility in their modernization efforts. The real-time pattern recognition the Cognitive Computing Toolkit offers will be of interest to any industry, though oil, advertising, and security will certainly find it of immense value. Best of all, HPE’s recently announced GPU-based HPE Apollo 6500 System will provide businesses with a high-performance computing (HPC) system powerful enough to handle these massive data streams.

So to all the developers out there who have always wanted to get in on the inception of a world-changing idea (or who have been there once and want to revisit that experience), who have the drive to influence the direction and quality of our journey to the future of computing: Download the Cognitive Computing Toolkit on GitHub and take your place at the sharp edge of cognitive computing.

For an in-person introduction to the Cognitive Computing Toolkit, visit our live demo at HPE Discover in Las Vegas.

Photo by Florian Richter via Flickr (CC BY 2.0)

0 Kudos
About the Author

Curt_Hopkins

Managing Editor, Hewlett Packard Labs