Google Launches TensorFlow As An Open Source Project

Posted: Nov 11 2015, 2:16am CST | by , Updated: Nov 11 2015, 2:19am CST, in News | Technology News

Google Launches TensorFlow as an Open Source Project

The next-generation machine learning system is used across Google's products, including Google Search, Google Inbox, and its Photos app.

Google is open-sourcing TensorFlow, the company's latest machine learning system that is used across its products and services, including Google Search and the newly released Photos app. An earlier version of deep learning technology, dubbed DistBelief, made it possible for engineers to build larger neural networks. However, configuring the system proved to be difficult, and sharing code externally is almost impossible, writes Jeff Dean, Senior Google Fellow.

TensorFlow was developed from the ground up to “right the wrong,” bringing improvements to DistBelief's speed, scalability, and production-readiness. The second-generation machine learning system is portable, flexible, and easy to use. But Google is upping the ante by making it an open source library for virtually anyone to use and develop. TensorFlow is twice as fast as DistBelief, according to Google's benchmarks.

Any computation that can be expressed as a computational flow graph can be computed using TensorFlow, says Rajat Monga, the Technical Lead of the project. Apart from using TensorFlow for research projects, it can be applied to actual products as well. It can run on datacenters, networks of computers, or even on a smartphone. Google, for instance, uses the technology to improve the speech recognition in Google Search, as well as on Smart Reply in Inbox. This allows Google to improve its products faster.

Google uses the C++ programming language to build TensorFlow's software. Developers, however, can use Python to create applications based on the engine. By open-sourcing the technology, Google is inviting software developers to help TensorFlow support other languages like JavaScript, Java, and Google Go. The open source library will be managed by Google at a new website at The code will be shared using an Apache 2 license. The move was lauded by the software development community.

“We've seen first-hand what TensorFlow can do, and we think it could make an even bigger impact outside Google. So today we’re also open-sourcing TensorFlow. We hope this will let the machine learning community—everyone from academic researchers, to engineers, to hobbyists—exchange ideas much more quickly, through working code rather than just research papers,” says Sundar Pichai, CEO of Google. “And that, in turn, will accelerate research on machine learning, in the end making technology work better for everyone.”

This story may contain affiliate links.


Find rare products online! Get the free Tracker App now.

Download the free Tracker app now to get in-stock alerts on Pomsies, Oculus Go, SNES Classic and more.

Latest News


The Author

<a href="/latest_stories/all/all/21" rel="author">Gene Ryan Briones</a>
Gene Ryan Briones (Google+) is a technology journalist with a wide experience in writing about the latest trends in the technology industry, ranging from mobile technology, gadgets and robots, as well as computer hardware and software.




comments powered by Disqus