ISSUE 317 ·   January 5, 2021        

In the News

Medicine's Machine Learning Problem

Data science offers the potential to fundamentally alter medicine but despite the promise, biased datasets and unaccountable algorithms threaten to further disempower patients. 
Boston Review | Rachel Thomas

Inside India’s booming dark data economy

In India, the black market for data resembles markets for wholesale vegetables and smuggled goods. Customers are encouraged to buy in bulk, and the variety of what’s available is mind-boggling.
Rest of World

Trends

Machine learning is going real-time

In her latest post, Chip Huyen explores the current state of real-time, online machine learning including use-cases, solutions, and challenges. After interviews with about a dozen companies, she concludes that "machine learning is going real-time, whether you’re ready or not."
Chip Huyen

Sponsored Link

2021: The year you make ML (and your job) easier

Make it easy to get your ML projects from experiment to production. Comet automatically tracks datasets, code changes, experimentation history and production models - all so you can focus on data science. Get started today with the free community edition.

Tutorials, Projects & Opinions

The value of p

What we lose if we abandon p values.
Richard D. Morey

NumPy Illustrated

This visual guide is a great introduction to NumPy and has gotten a lot of attention around the web recently. It builds on a previous article by Jay Alammar and covers a broad range of operations including vectors, matrices, and high dimensional operations.
Lev Maximov

Code & Tools

Top 10 Python libraries of 2020

Tryolabs' annual list of top Python libraries is consistently a must-read post. This year's list is tailored for data science and ML and includes tools for high-dimensional plotting, config management, forecasting, command line interfaces, productivity, outlier detection, and more.
Tryolabs

Uncertainty Toolbox

This Python toolkit provides standard metrics to quantify and compare uncertainty estimates from ML methods. It  also offers intuition for these metrics, produces visualizations, and implements simple "re-calibration" procedures to help improve the uncertainties. This is well-documented and includes a collection of linked references.
GitHub | uncertainty-toolbox

JupyterLab 3.0 is released!

Jupyter 3.0 is a major release that includes some key features including a visual debugger, a table of content for notebooks, multiple display languages, and a much improved extension system. See this announcement post for details.
Jupyter | Jeremy Tuloup

Resources

2020: A Year Full of Amazing AI papers- A Review

This curated list of the latest breakthroughs in AI include short video explanations, links to in-depth articles, and code.
GitHub | Louis-François Bouchard

⚽ Analytics 2020 Review

Awesome roundup of ⚽ analytics content from 2020. Covers research papers, blog posts, podcasts, talks, Python libraries, datasets, and more!
Jan Van Haaren

Data Elixir is curated and maintained by Lon Riesberg. For full-text search of prior issues, visit Data Elixir's Search Page. If you have suggestions or questions for the newsletter, just reply back to this email.

Sign up to get Data Elixir's  data science newsletter in your Inbox >>