I would rather say in addition to numpy. Tensorflow, and the more recent CNTK may be good candidates for a production environment. Keras is a simplified interfa

I would rather say in addition to numpy. Tensorflow, and the more recent CNTK may be good candidates for a production environment. Keras is a simplified interface to those. Other libraries that people have used include PyTorch and Scikit-learn. But right now Tensorflow and CNTK might be among the fastest/more scalable.