Google Updates Distributed Computing To Its TensorFlow Machine Learning Models

By Samir Ghanmi | Apr 14, 2016 10:06 AM EDT

TEXT SIZE    

Google announced an update to its open-source framework TensorFlow that will now run training process for creating machine learning models over hundreds of machines aligned.

According to The Verge, Google opened up its TensorFlow last year for companies that wants to build their own artificial intelligence application using the same open-source library the search engine applies to power everything from photo analytics and automated email replies.

Google tried extending its platform to other computer servers by publicly releasing a version that could only run a single machine.

Now, Google has updated a new version of TensorFlow with a feature that will enable to run distributed computing across multiple machines at the same time. Engineering leader of TensorFlow Rajat Monga said the reason why TensorFlow's multi-server version was delayed for release because they found it difficult to adapt the open-source software to be usable outside of the highly customized data centers of Google. "Our software stack is differently internally from what people externally use. It would have been extremely difficult to just take that and make it open source." He said.

Google chose to release last year's limited version of the feature so that researchers and companies would have at least something to work with while the TensorFlow team worked on a more advanced feature.

TensorFlow comes in a branch of artificial intelligence called deep learning, it works the same way human brain cells interact together. Deep learning has become the pivot for providing machine learning to giant tech companies such as Facebook, Microsoft and Yahoo.

Regardless of the advanced feature, TensorFlow has already gained popularity for its software. Since it was released in November, TensorFlow was one among other six open-source projects that gained most recognition from developers. 

pre post  |  next post
More Sections