Luckily, for all major operating systems—that is, Windows, Mac, and Linux—there are targeted installers for NumPy, SciPy, Matplotlib, and TensorFlow. If you are unsure about the installation process, you might want to install the Anaconda Python distribution (which you can access at https://www.anaconda.com/download), which is maintained and developed by Travis Oliphant, a founding contributor of SciPy. Luckily, Anaconda is already fully compatible with Python 3—the Python version we will be using throughout this book.
The main Anaconda channel comes with three flavors of TensorFlow (use the Intel channel at your own risk, that is an older version of TensorFlow). The main flavor, tensorflow, is compiled for all platforms and runs on the CPU. If you have a Haswell CPU or a more recent Intel one, you can use the tensorflow-mkl package. Finally, if you have an Nvidia GPU with a compute capability of 3.0 or higher, you can use tensorflow-gpu.