TensorFlow is the Deep Learning library with the most growth potential
This article was originally posted at https://infosimples.com/ by Rafael Barbolo on November, 2015. It has been imported to Infosimples Medium, the new home of Infosimples articles.
Deep Learning libraries have gained popularity in Machine Learning applications. Here are the leading ones:
Library Support Top users Caffe C++, Python NVIDIA, general industry SINGA C++, Python Apache Theano Python Yoshua Bengio (University of Montreal),
some researchers at Google Torch C, Lua Yann LeCun (Facebook)
I’ll not get into details on each one. It’s worth mentioning that Theano and Torch are mostly used in cientific research, Caffe is mostly used in commercial Machine Learning applications and SINGA was launched recently by the Apache Foundation and seems to compete mostly with Caffe.
At Infosimples we mostly use Caffe because it has great support for processing images. Any of these libraries is a great choice and they have great communities with active development.
A new options was introduced yesterday and that’s why I’ve decided to write this mini article. I’m talking about TensorFlow, developed by a group of renowned researchers at Google. The researchers themselves are a top reason to call our attention. They have published great articles in Machine Learning conferences and some of them have contributed to the development of the libraries I mentioned in the table above.
In addition to being flexible to be used by the cientific community, TensorFlow is also a great choice to deploy commercial applications of Machine Learning, with great support to run in servers, desktops and mobile devices.
I think this tool have everything in place to become a leading resource in Machine Learning development. So, if you want to get into this field, now is a great time. I recommend you to begin with the tutorials of TensorFlow.
I hope to bring more news soon. I’ll leave the TensorFlow launch video below.
Originally published at infosimples.com.