“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.
Advertisement
CNTK was compared in the internal tests with Torch 7, Theano and Caffe, as well as Google’s TensorFlow that was recently open-sourced.
Chris Basoglu, a principal development manager at Microsoft who also worked on the toolkit, said one of CNTK’s advantages 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. For example, the company’s popular virtual assistant Cortana takes advantage of CNTK for speech recognition.
“We further introduce the computational network toolkit (CNTK), an implementation of CN that supports both GPU and CPU”.
CNTK first went open source in April 2015, but was hosted on Microsoft’s CodePlex repository, where users had to adhere to Microsoft’s academic license.
But now that it’s freely available, they believe it could be useful to anyone from deep learning startups to more established companies that are processing a lot of data in real time.
Huang and his team developed the toolkit out of necessity: They wanted to improve how well computers can understand speech, but all the tools they had were slowing them down. According to Venture Beat, Microsoft Corporation is also dropping the company’s research license, which limits the software to non-commercial uses, in favor of a more permissive MIT license.
Microsoft wants to make its technology available for other companies and academics to take advantage of. Well, Microsoft is betting big on deep learning and artificial intelligence, so its researchers chose to develop their own modern toolkit called Computational Network Toolkit (CNTK), in hope of reaching more breakthroughs in the machine learning field. “That strengthens the ecosystem and the toolset we have”, Huang tells VentureBeat. “Many researchers see deep learning as a very promising approach for making artificial intelligence better”, Microsoft said.
According to MIT Technology Review, the speeches consisted of more than 50,000 sentences each containing 23 words on average.
Advertisement
“Mr. Speaker, for years, honest but unfortunate consumers have had the ability to plead their case to come under bankruptcy protection and have their reasonable and valid debts discharged”. Stand up for growth and opportunity.
