As software becomes more intelligent by the day, Machine Learning platforms have become mainstream. ML platforms have come from university research labs as well as leading organizations that have employed them in their products. One area that we've seen the predictive power of these systems has been in web search, with Google leading the charge. However, it has been frustrating to see that while several vendors had announced ML platforms, Google at best released its Prediction API that had not seen an update in quite a while. That is no longer the case, with the announcement of TensorFlow, a battle tested Deep Learning platform, that is now available to everyone.
Jeff Dean, Senior Google Fellow announced TensorFlow at the Google Research Blog. the post gives a bit of history about TensorFlow, which they call a second-generation platform, building on the base of DistBelief, their internal deep learning infrastructure that was built in 2011. TensorFlow is already powering multiple applications at Google such as the Google Speech Recognition API and Google Photos.
At its heart, TensorFlow is a library that you can plug into your applications for numerical computation using data flow graphs. As per the blog post, the platform is general enough that you can apply it to a wide range of cases.
The most noteworthy thing about this announcement is that Google has Open Sourced the entire TensorFlow platform and is banking on developers to take it to the next level with enhancements, multiple language bindings and more.
The current programming mode is via Python and C++ interfaces but gauging from the interest that the release has generated among developers, we can soon expect other language support. The TensorFlow site is complete with excellent documentation that includes Getting Started with the platform. The platform can be downloaded and setup on your systems in multiple ways, including a binary distribution, a ready to go Docker image and more.
Google’s move to release such important technology as open source as compared to other major vendors like Facebook and Microsoft is commendable and goes to show the confidence they have in the community to adopt this and take it ahead to help make software more intelligent in the days to come.
Check out the TensorFlow site for more details.