What is TensorFlow?
Google just announced the release of TensorFlow, a new, open-source machine learning system. It is a second-generation system that was developed from the successful DistBelief deep learning infrastructure, and features improved speed, scalability, and production readiness.
DistBelief was useful in a number of applications. For example, it improved speech recognition in the Google app by 25% and was also used to label uncategorized images (it could identify images of cats, for example, and label them accordingly). However, the DistBelief code was tied to Google and couldn't be easily exported for other users. TensorFlow changes all of that.
Jeff Dean, Senior Fellow in the Systems and Infrastructure Group, describes the system on the Google Research Blog as “general, flexible, portable, easy-to-use, and completely open source.”
The New Age of Open-Source
TensorFlow will have extensive built-in support for deep learning, which makes it able to compute anything that can be expressed as a computational flow graph. It also features auto-differentiation and a suite of first-rate optimizers, which are a fantastic benefit to any gradient-based machine learning algorithm.
Built to be fast, portable, and a ready-for-production service, TensorFlow can be easily implemented in a number of products. Dean notes, "You can move your idea seamlessly from training on your desktop GPU to running on your mobile phone."
Ultimately, Dean highlights open source as the TensorFlow’s main feature, saying that “the most important thing about TensorFlow is that it’s yours. We’ve open-sourced TensorFlow as a standalone library and associated tools, tutorials, and examples with the Apache 2.0 license so you’re free to use TensorFlow at your institution (no matter where you work).”