Our Blogs

Google TensorFlow 1.0 gets better with new machine learning tools

February 16, 2017
Google has announced the launch of TensorFlow 1.0, which is an open-source framework for deep learning platform. Touted as a trendy fashional type of artificial intelligence, the TensorFlow 1.0 is currently ready for production and can be consumed via its Application Programing Interface (API).

Commenting on the development, Rajat Monga, TensorFlow's engineering director disclosed that there are new tools that are released as part of the new framework. It includes artificial neural networks, which will have an ability to consume data. Currently, we have access to a wide range of machine learning tools such as K-means and support vector machines (SVMs).

TensorFlow 1.0 ships with deep integration with the Python-based Keras library. The main purpose of the library is to simplify the use of the Theano deep learning framework. Google is also planning to open-source the coding framework. You can expect improved speed of TensorFlow after that. Furthermore, TensorFlow will be integrated with support for the Hexagon digital signal processor (DS) , which comes with Snapdragon 820 mobile chip and its Dragonboard 820c board. 

Google also added an experimental TensorFlow compiler (XLA) to the TensorFlow just-in-time (JIT) compiler. It will automatically compile a graph leading the path to assembly language. Moreover, you have access to an experimental Java API paired with a debugger.

Originally launched in 2015, the TensorFlow platform has been frequently refreshed with  more features. Some of the features are distributed training, Hadoop Distributed File System (HDFS), and the Parsey McParseFace language parser.

The search-engine giant also offers the Cloud Machine Learning service that helps you to run TensorFlow on Google's cloud infrastructure. Hence, you will be able to use the platform for your custom software development requirements.