Microsoft launches open-source Cognitive Toolkit 2.0

June 4, 2017, 8:37 a.m. By: Vishakha Jha

Cognitive Toolkit 2.0

Microsoft Cognitive Toolkit version 2.0, previously known as CNTK is now available under an open-source license as announced by Microsoft on June 1. Since the beta release of the Cognitive Toolkit in October 2016, the technology has been accepted by companies across the world to define and train neural networks, which are systems that can learn how to perform a task similar to human brain.

The Microsoft Cognitive Toolkit is a deep-learning toolkit which describes neural networks through a directed graph with leaf nodes representing input values and other nodes represent matrix operations upon their inputs.

Cognitive Toolkit 2.0 includes many new features through deep learning algorithm which will allow developers to create and evaluate their own neural networks that can scale across GPUs and other machines on massive data sets and it has also improved search relevance. It is built with a sophisticated algorithm that offers expressive and easy-to-use architecture.

The Toolkit supports Keras, a user-friendly open-source neural network library used for deep learning applications. The Keras API takes a user-centric approach to developing AI-enabled applications that could help the user with little experience on the subject to get results. It enhances training capabilities by including Java language bindings used for evaluating models and supporting the latest versions of NVIDIA’s Deep Learning SDK and advanced GPU’s.

There are also model compression extensions enabling image recognition and other Cognitive Toolkit models to run faster on standard servers, smartphones and embedded devices with more modest processing capabilities.

Although Skype, Cortana, Bing, Xbox and leading data scientists already use the Cognitive Toolkit but now it is part of an initiative to make AI technology accessible to everyone. In addition to the Cognitive Toolkit, developers can access a suite of cloud computing applications via Microsoft Azure or APIs, via Microsoft Cognitive Services.

Image Source: WinBuzzer