Google LLC today introduced several major improvements to TensorFlow that significantly extend the machine learning development framework’s capabilities.

The first pair of enhancements aim to grow the number of areas where the platform can be applied. Google has added a library for processing genomics file formats, which should help expand the use of TensorFlow in cutting-edge medical projects that rely on DNA data, and implemented a set of features to ease so-called Bayesian analysis.

The latter term refers to a statistical approach commonly used in situations where there’s not particularly much data with which to train an artificial intelligence. Thanks to the new Bayesian building blocks, TensorFlow should reduce the amount of work required to produce a working model under such circumstances. In the same spirit, Google has also added “increased support” for a number of more traditional machine learning methods.

Many of the other improvements similarly focus on streamlining the AI development workflow. TensorFlow is receiving a programming model dubbed eager execution that aims to simplify coding, plus a library called TensorFlow Hub for sharing of commonly used AI elements. The idea is to remove the need for developers to reinvent the wheel and free them up to focus on the more unique parts of their projects. 

TensorFlow Hub is joined by two tools that Google has created to help engineers find potential issues with their neural networks faster. The first is a graphical debugger for inspecting the artificial neurons of an AI, while the other is designed to visualize how well the model as a whole analyzes large amounts of data.

Completing today’s update is support for a slew of new third-party technologies. On the hardware side, TensorFlow now works with the Raspberry Pi miniature computer and Nvidia Corp.’s TensorRT, a framework for optimizing AI models to run better on its market-leading graphics processing unit chips.

On the software front, Google is adding support for two additional programming languages: JavaScript and Apple Inc.’s Swift. The former edition is particularly significant because the search giant said it will pave a path to the development of AI algorithms capable of running in a browser.

Image: Google