Microsoft posts AI toolkit on GitHub

The move to fully open-source the CNTK tools is part of Microsoft's effort to gain mind share and boost Azure pickup

Microsoft posts AI toolkit on GitHub

Seeking to spread usage of its deep learning software, Microsoft is making its open source CNTK (Computational Network Toolkit) for artificial intelligence available on GitHub.

The kit, built with an emphasis on performance, provides a unified computational network framework describing deep neural networks as a series of computational steps via a directed graph, according to Microsoft Research.

"The researchers developed the open-source toolkit, dubbed CNTK, out of necessity," according to a Microsoft blog post Monday by Allison Linn. "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." Internal tests found CNTK more efficient than other popular computational toolkits, including Theano, TensorFlow, Torch 7, and Caffe, Microsoft said.

CNTK has been available via open source since April 2015, but had been offered on Microsoft's CodePlex site under a more-restrictive license. Now, it's available to anyone with an MIT license.

This open source move is continuing Microsoft's effort to demonstrate that it's now more community-focused, analyst Rob Sanfilippo, of Directions on Microsoft, said.   "This strategy is geared to gain mind share for the company for solutions where its paid offerings have not been adopted due to licensing requirements or the opacity of the software. Open source software is a trend that can no longer be avoided by Microsoft and the majority of industry members."

The effort, Sanfilippo said, could pay off by sparking interest in other Microsoft services, such as Azure. "Organizations with developers working on speech and image recognition, and deep learning and neural networks in particular, will likely be interested in evaluating CNTK since the performance gains could speed projects, especially where multi-GPU hardware is available."

Microsoft has used CNTK on a set of computers using GPUs, which have been found useful in processing algorithms leading to advances in technology that can speak, hear, and understand speech and recognize images and movements. The company plans to leverage CNTK and the Azure GPU Lab to provide a distributed GPU platform for use in advanced AI research. In recent years, the field of deep learning has grown with more researchers running machine learning algorithms using deep neural networks. Researchers see deep learning as an approach to improving artificial intelligence, Linn said.