Microsoft Cognitive Toolkit 2.0 Beta With Python Support Launched

Microsoft has announced the launch of beta version 2.0 of its open-source deep learning framework, Microsoft Cognitive Toolkit. Till now, the software was known as Computational Network Toolkit or CNTK. Now, the company is changing its name to Cognitive Toolkit because the new name lines up more clearly with Microsoft’s “Cognitive” marketing push of the past few months.

Instead of just being accessible through the programming language C++, Cognitive Toolkit now comes with native support for Python, especially for Python 3. While, the Python 2 support is coming.

Performance has also been improved along with reinforcement learning, an approach involved with the teaching machines through trial and error, which has been added to the 2.0 release. You will now be able to easily work with this software in Microsoft’s Visual Studio integrated development environment (IDE) as opposed to just running Cognitive Toolkit as a script.

Deep learning involves training artificial neutral networks on several data, such as images and then getting them to make inferences about the new data, along with several other open-source options exit.

Google is currently using deep learning, especially TensorFlow, at several places, and it seems that Microsoft is also doing the same with Cognitive Toolkit. Products such as Windows, Bing, and Skype Translator, along with others, have been using it in different ways. Now, with the latest version 2.0, Microsoft has come up with something that it believes several other companies can also put into production.

Cognitive Toolkit is general-purpose software; hence it will be working well on multiple kinds of data, which include speech, text, and images. The training phase can now be done on multiple standard X86 Central Processing Units (CPUs) or Graphics Processing Units (GPUs). The distributed aspect is really important as it would go on to enable faster computing.

With time, Microsoft wants to make Cognitive Toolkit available more widely with the support for R and C#, however, for now you can try the Cognitive Toolkit on the GPU backed instances offered by the Microsoft Azure public cloud. The cloud is open in the sense that it can now easily go on to support TensorFlow and Caffe.

The new benchmarks from Microsoft show Cognitive Toolkit outperforming TensorFlow Torch and Caffe for training of two particular scenarios – the f"our GPUs and one CPU" and "four GPUs and two CPUs".

For more information, check the official blog.