Facebook has reportedly shared some of their technologies as a part of their open source initiatives. It has been announced by Facebook's artificial intelligence research team that their deep-learning AI tools are getting open sourced.



The software has been made available via Torch library, which is an open source environment for development of machine learning. Torch is a widely used medium for research in academia and it's used by companies like Google, Twitter and Intel. 
Facebook claim says that their deep-learning modules are much faster than the default ones which are available on Torch. These tools allow the company work on larger neural networks in much lesser time.

 The improvements in the modules include 23.5x speed in the layer codes and also their ability to work with parallel neural networks while training over GPU cards. The company has shared their modules with the hope that they will optimise progress across the deep-learning scenario. Facebook wants that the entire world sees more patterns and predictions. Deep-learning is a kind of machine learning which copies how scientists think how brain works and it separates meaningless information from meaningful signals.

 This move by Facebook has self-interest too. Now Facebook will be able to accelerate their own AI projects and such open source techniques have been reveled previously by Mark Zuckerberg as a part of his Open Compute Initiative. This initiative was launched to compete with Google, Amazon and Yahoo to build big data centers. 

 Facebook artificial intelligence researcher and software engineer Soumith Chintala says the open source project will help research labs and startups which lack from resources and implement existing algorithms instead of doing new research. Facebook uses deep learning in filtering Facebook feed and to recognise faces in uploaded photos. 

You can learn more about facebook AI at following link:
Facebook AI Research

TAGS :