— In the News —
Google made a huge splash this week by open-sourcing their machine learning system called "TensorFlow." This is a BIG deal. Check out the announcement on the Google Research Blog and then follow up with these:
Research that can't be reproduced is hard to trust but doing things "right" is complicated by the ubiquity of personal computers and data that's hard to share. This is a must-read article about the importance of reproducibility in science, how that's failing, and ideas for getting back on track.
Facebook's new AI, called M, is said to have capabilities that far exceed those of competing AIs. Some people claim that it's actually human-assisted but M insists it's an AI. A Turing Test won't work here because M’s objective is precisely to not pass a Turing test. This is a fun exploration of how to test what's really going on behind the scenes.
— Tools and Techniques —
Second part of a two-part series that describes lesser known ways of using Jupyter notebooks. This week's article shows how to use widgets to build interactive dashboards. In case you missed it, Part 1 shows how to create pipelines and reports.
One of the best things about the R ecosystem is being able to rely on other packages so that you don't have to write everything from scratch. But how do you know which packages can be trusted? This post by Jeef Leek offers a worthwhile checklist of considerations.
— Data Viz —
Fantastic collection of strategies for presenting time-oriented data. There are 113 techniques here that are easy to browse using multiple filters and/or keyword search. Each technique includes a screenshot, description, and linked references.
Nice overview of commonly used Python data visualization tools. This post by Vik Paruchuri uses a real-world dataset to explore the capabilities of matplotlib, vispy, bokeh, seaborn, pygal, folium, and networkx.