Microsoft took another step on its open-source sharing journey Monday by releasing on GitHub a toolkit it uses internally for deep learning.
Dubbed CNTK -- short for Computational Network Toolkit -- the open-source software is a unified deep-learning toolkit that describes neural networks as a series of computational steps via a directed graph. It's proven significantly more efficient than Google's TensorFlow tool and other competing tools, Microsoft says.
Microsoft has used CNTK internally to help advance speech recognition in products such as Cortana, but it says it could be useful to a wide variety of other users as well, including not just companies focused on deep learning but also those that need to process large quantities of data in real time.
The software has been available to academic researchers for almost a year under a more restrictive custom license via Microsoft's CodePlex site. Now on GitHub, it's available for download using the more permissive MIT license.
Last November, Microsoft made much the same move with its Distributed Machine Learning Toolkit (DMTK). That same month, Google took a similar step by releasing its own TensorFlow machine-learning system under an Apache 2.0 license.
Microsoft has been using CNTK on a set of GPU-based computers, but such computational muscle isn't required, it says.
"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," explained Xuedong Huang, Microsoft's chief speech scientist, in a December blog post. "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."