Microsoft releases machine learning toolkit on GitHub in another win for AI research

Tools

Microsoft today announced it released another set of machine learning tools to GitHub. The company claims the tools are more efficient than its competitors'. 

The space has become contentious in recent months as Microsoft, Google and IBM all made more widely available similar artificial intelligence technology to open source developers in the span of a few weeks near the end of last year.

It's become something of a machine learning advent for developers outside of the "tech giant" bracket who, until now, had limited options for testing AI capabilities. It seems that as the technology sees more use across those large companies' businesses, they're working to generate interest among developers who could build out capabilities on the platforms.

Microsoft had previously made its Computational Network Toolkit, or CNTK, available for academics under its own project hosting website CodePlex. The company wrote previously that it used insights gleaned from CNTK in its own Cortana, Skype Translator and Project Oxford Speech API technologies.

However, this move to GitHub will open it up to a larger set of developers.

"The CNTK toolkit is just insanely more efficient than anything we have ever seen," Xuedong Huang, chief speech scientist at Microsoft, said in a blog announcing the GitHub release.

Huang, writing previously on the subject in a Microsoft Research blog, outlined how the CNTK offering was more efficient than competitors like – Google's TensorFlow, Theano, Torch 7 and Caffe – when paired with Microsoft's Azure GPU Lab. He further wrote that while each offering has "unique strengths," he thought Microsoft took home the win on brute strength.


Source: Technet Blog

"TensorFlow offers a user-friendly Python interface; Theano is unique with its symbolic operation; Torch uses Lua programming language; Caffe is popular for computer vision researchers due to its efficient performance; and CNTK on Azure GPU Lab offers the most efficient distributed computational performance," Huang said.

Regardless of which technology excels, getting developers up to speed on machine learning and its associated technology seems to be the goal here. As Microsoft dives deeper on AI and the products and services it can provide, the company would do well to cultivate developers that can build for those kinds of systems.

For more:
- read the blog announcement today from Microsoft
- read the previous blog post from Huang on CNTK capabilities

Related Articles:
Apache Incubator accepts IBM's SystemML for open source development
Microsoft releases machine learning toolkit DMTK to open source
Google shares its machine learning software, TensorFlow, with the world