Google announced that it will open source (Apache 2 license) a single machine (CPU with multiple GPU’s permitted) of its Machine learning software.
The software already underpins its Photos, image recognition, speech recognition in Android, Chrome, and Gmail responses, Google Translate, among other applications.
For the few that want to learn, it is available now.
This will dramatically change neural networks and intelligence overall I believe.
Until now, this was available via Azure ML from Microsoft and AWS Amazon Machine Learning, but they are both not open source.
With this new software, Google allows any software developer who wants to use Python or C++ as interfaces to train the network on their own data.
The typical use cases I see are when developers want to “try” things in their own machine and “tinker” at first.
The applications to use Tensor flow are pretty extensive.
Many SaaS companies wanting to do predictive Churn analysis can now do this internally, as can eCommerce companies to build recommendation engines as well.
If you are a developer I’d highly recommend you download the PIP and install it on a Linux distribution on your local machine. Or if you have a Mac, Homebrew and install.