by bill-s, 2019-01-22T09:19:08.506Z
It’s possible, though quite difficult, to create neural networks from raw code. Luckily, there are many open source code libraries you can use to speed up the process. These libraries include CNTK (Microsoft), TensorFlow (Google) and scikit-learn. Most neural network libraries are written in C++ for performance but have a Python API for convenience.
In this article I’ll show you how to get started with the popular PyTorch library. Compared to many other neural network libraries, PyTorch operates at a lower level of abstraction. This gives you more control over your code and allows you to customize more easily, at the expense of having to write additional code.