5 important Python Data Science advancements of 2015

The post says it is hard to keep up with the pace of change in the Python data science community, so it offers a short recap of the most notable developments from PyData and SciPy talks. It focuses on tools and workflows that made Python faster, easier to manage, and easier to share.

Reading notes

#

• Numba brings just-in-time compiling, so Python code can get significant speed-ups without rewriting the whole program or pre-compiling it.
• Conda is presented as a package and environment manager that is better than pip for packages, better than virtualenv for environments, and platform independent.
• The Jupyter notebook has become a broader ecosystem after separating from the iPython notebook kernel, with support across languages and related tools like GitHub rendering, Binder, tmpnb.org, and Pineapple.
• GitHub now renders .ipynb files directly, which helps make notebook-based analysis easier to share.
• Binder can turn a GitHub repository into interactive notebooks in the cloud, which the post says helps sharing and reproducibility.
• tmpnb.org can start a temporary notebook instantly.
• Pineapple is a desktop app for Mac, though the post says it still needs work.
• Jupyter is expected to get support for real-time collaboration.
• Matplotlib 2.0 is described as moving toward a prettier default style, while still allowing styles such as ggplot, bmh, and fivethirtyeight.
• The post says matplotlib 2.0 will break backward compatibility, and recommends using Seaborn as the interface.
• TensorFlow is described as a second-generation deep learning library from Google that quickly gained more followers than other open source deep learning libraries on GitHub.