Microsoft has broadened open-source access to its Computational Network Toolkit, in a move that underscores the ongoing arms race around machine learning technology.
CNTK had initially been made available on Microsoft's Codeplex site in April 2015 through under an academic license, but now it has been moved to GitHub, one of the web's largest code hosting sites. Microsoft has also switched CNTK to the much more permissive MIT license.
By making these moves, Microsoft is hoping to build up a broader community of developers around CNTK. The company provided some background on CNTK in an official blog post announcing the changes:
Xuedong Huang, Microsoft's chief speech scientist, said he and his team were anxious to make faster improvements to how well computers can understand speech, and the tools they had to work with were slowing them down.
So, a group of volunteers set out to solve this problem on their own, using a homegrown solution that stressed performance over all else.
In internal tests, Huang said CNTK has proved more efficient than four other popular computational toolkits that developers use to create deep learning models for things like speech and image recognition, because it has better communication capabilities
"The CNTK toolkit is just insanely more efficient than anything we have ever seen," Huang said.
Those types of performance gains are incredibly important in the fast-moving field of deep learning, because some of the biggest deep learning tasks can take weeks to finish.
Chris Basoglu, a principal development manager at Microsoft who also worked on the toolkit, said one of the advantages of CNTK is that it can be used by anyone from a researcher on a limited budget, with a single computer, to someone who has the ability to create their own large cluster of GPU-based computers.
The researchers say it can scale across more GPU-based machines than other publicly available toolkits, providing a key advantage for users who want to do large-scale experiments or calculations.
Those "other publicly available toolkits" include Google's TensorFlow, which was open-sourced last year. IBM, Baidu and others have also open-sourced their own machine learning toolkits.
Since the initial release of CNTK, Microsoft has made significant advancements, company researchers said in a December blog post:
The combination of CNTK and Azure GPU Lab allows us to build and train deep neural nets for Cortana speech recognition up to 10 times faster than our previous deep learning system. Our Microsoft colleagues also have used CNTK to run other tasks, such as ImageNet classification and a deep structured semantic model. We've seen firsthand the kind of performance CNTK can deliver, and we think it could make an even greater impact within the broader machine learning and AI community. It's our hope that the community will take advantage of CNTK to share ideas more quickly through the exchange of open source working code.
Microsoft isn't betting its entire machine learning research agenda on CNTK, however. Indeed, in November it released another project, Distributed Machine Learning Toolkit, to Github.
Analysis: Microsoft's Moves Reflect the Mounting AI Arms Race
"Deep Learning is a cutting-edge area of machine learning that in recent years has brought about rapid advances in speed recognition, voice search and pattern recognition used in areas such as machine vision," says Constellation Research VP and principal analyst Doug Henschen. "A key difference with deep learning is its emphasis on unsupervised or semi-human supervised analysis. In short, it's a building block of broader research on Artificial Intelligence (AI), and Microsoft is clearly attempting to raise its profile and advance its work through the open-source community model."
The move to GitHub will greatly increase CNTK's exposure and distribution, Henschen adds. In addition, the MIT license will allow developers to use the code in combination with proprietary software, he notes.
Meanwhile, processing speed and throughput are the name of the game in machine learning," Henschen says. "The faster you can process the data, the more data you can use and the more quickly the machines can learn. Using more data brings greater accuracy to machine learning."
"With its ability to run on a single, high-end Graphical Processing Unit (GPU) server or distributed clusters of GPU-based machines--a differentiator, according to Microsoft--CNTK appears to give developers speed and throughput advantages," he adds. "Of course, research in machine learning is fast moving, so I'd expect plenty of competition to keep the bar moving higher. The point here is that Microsoft is stepping up in a very public way to promote its research on machine learning."
Constellation Insights is an online news service published daily to advise members of the Insights community on the significance and implications of developments in enterprise technology.
Constellation Insights is crafted by leading analysts to go beyond merely reporting on news stories to provide detailed advice to community members on how to evaluate and respond to changes in enterprise technology. Learn more about Constellation Insights.