— In the News —
With all the COVID-19 models floating around, how does anyone know what to expect? These guidelines from John Hopkins University offer a nice framework for thinking through the various models and evaluating what they can and can't do.
Facebook has open-sourced a new chatbot, called Blender, that's said to be the best yet. With 9.4 billion parameters, it may not be practical to use yet but its capabilities are impressive.
— Tools and Techniques —
This collection of SQL problems is designed for people who already work with SQL and want to level-up their skills. Includes a variety of problem types, solutions, and linked references.
Getting ML into production is hard. There are MLOps solutions that help solve the process of building and deploying models but they lack support for one of the most challenging parts of ML: the data. This post explores the issues and how a new player called Tecton aims to help.
How does the New York Times responsibly take advantage of machine learning? It helps sort reader comments, but humans do the moderating.
Great post if you're just starting out or thinking about expanding your current setup. Includes lots of detail, links to key resources, and rationale for decisions along the way.
Want to become a better Pythonista? PyCoder's is a free, weekly email newsletter for those interested in Python development and various topics around Python and the community.
— Resources —
To help support researchers and students through the COVID-19 lockdown, Springer has made more than 400 of its essential texts available for free download, including these CS books on Python, deep learning, data science and AI.
Papers and workshops from ICLR2020 are now available on the conference website. Unlike many of the tech conferences that cancelled this year, ICLR2020 made a successful transition to virtual. Check out the retrospective to see how they pulled it off.
— Data Viz —
A recent COVID paper used an unusual visualization that got a bit of attention around the web last week. Martin Wattenberg dubbed it a "tornado plot" and followed up with this interactive post to help make sense of the technique. It certainly is attention-grabbing. But useful?