Open source sourcing "TensorFlow", a machine learning system Google uses in product development, for commercial use OK

A system for deep learning and machine learning that Google uses also in its product development "TensorFlowOpen source, and released it as an Apache 2.0 license. Now that the machine learning system Google uses can be freely used by rival companies, Google seems to have liberalized in anticipation of the future.

Home - TensorFlow
http://tensorflow.org/

Research Blog: TensorFlow - Google's latest machine learning system, open sourced for everyone
http://googleresearch.blogspot.jp/2015/11/tensorflow-googles-latest-machine_9.html

Official Google Blog: TensorFlow: smarter machine learning, for everyone
https://googleblog.blogspot.jp/2015/11/tensorflow-smarter-machine-learning-for.html

Google announced in December 2011 a system for deep learningDist BeliefWe have been using it in the Google services group after that. Specifically, it automatically generates thumbnails of YouTube movies, improves accuracy of voice search,Google PhotosIt is that DistBelief's deep learning and machine learning have supported the high quality service of Google such as automatic tagging function of Google tag translation and Google translation from behind.

Developed that DistBelief is the 2nd generation machine learning system "TensorFlow", which solves the shortcomings of DistBelief and is realizing twice as fast on the benchmark. Furthermore, TensorFlow is designed to have flexibility to run from mobile terminals to thousands of connected computers despite its high speed, "If TensorFlow can be expressed as a computer program in data flow format, We can say that we can do it, "Google says that it can be used by members of every machine learning community, such as researchers, companies, users working on machine learning as a hobby.

A data flow graph depicting the program flow in TensorsFlow looks like this.

TensorsFlowing : check out http://tensorflow.org - YouTube


It is no doubt that TensorFlow is one of the key technologies to support the foundation of service development that Google has been developing for many years, why Google is going to open source, furthermore it can be used commercially in the service Apache 2.0 It seems strange whether it will be provided as a license. Google explained about why, "Machine learning is a key technology that is key to the creation of innovative products and technologies in the future, development is actively carried out around the world, but lacks standard tools I want to foster TensorFlow as an open standard tool that enables us to exchange research results and make machine learning easier for products by sharing TensorFlow that Google believes is one of the best machine learning tools in the world "It seems that we believe that providing machine learning tools will spread research on machine learning and create a new computing development environment will benefit Google.