ISSUE 367  ·   December 21, 2021

Note that Data Elixir is taking next week off and will be back in your Inbox on January 4th. In the meantime, stay safe out there and I hope you have great holidays! 🎉

-Lon

Insight

Do large language models understand us?

It’s sometimes claimed that machine learning is "just stats" and that AI can't really understand. But what does it really mean to "understand?" This is a great essay that will stretch your thinking about language, intelligence, understanding, sociality, and even personhood. 
Blaise Aguera y Arcas

Sponsored Link

Start working towards your career in data science.

Springboard’s Data Science Career Track is designed to help you master job-ready skills and actually land your dream job. With 1-on-1 mentorship, career coaching, and personalized support, you’ll gain the portfolio, skills, and confidence you need to get hired in a new role. Their students aren’t just finding new jobs, they’re launching new careers. Apply today.

Tutorials, Projects & Opinions

Reduced, Reused and Recycled: The Life of a Dataset in Machine Learning Research

Benchmark datasets play a key role in the organization of machine learning research but more and more, widely-used datasets are introduced by only a handful of elite institutions. This Best Paper winner from the recent NeurIPS Conference explores how that creates practical and ethical problems in machine learning research.
arXiv

Jupyter Games

Jupyter isn't usually thought of as an application for games but games are a great vehicle for learning how to push its edges. This tutorial shows how to build tiny games in Jupyter using the Box2D physics engine and Ipycanvas. Covers games like Rocket, Billiards, ~Angry Birds, and World of Goo — each in under 1000 lines of code.
Jupyter | Thorsten Beier

The Guide to Data Versioning

There are a lot of reasons you might want to version your data. If you're already versioning your code with git, here's how data versioning works using the same abstractions.
Whispering Data | Paul Singman

How to make AI & BI work at scale

Why isn't there more automation in data labeling? It makes sense to want to automate the labeling process as much as possible. Using human judgments seems expensive and inefficient. Industry experts weigh in on the pros and cons of applying automation in data labeling for machine learning.
// sponsored

Code & Tools

Top Python libraries of 2021

Tryolabs' annual list of top Python libraries is consistently a must-read post. This well-researched list includes tools for working with awkward arrays, versioning Jupyter Notebooks, monitoring ML models, working with time-series, interacting with SQL databases, and much more.
Tryolabs | Alan Descoins

Sharing Data With the pins Package

The pins package publishes R objects, such as datasets and predictive models, on a virtual cork board so you can safely share and reuse them. In other words, pins makes it easy to share and update data across projects and people.
RStudio Blog | Katie Masiello and Isabella Velásquez

Resources

From R to Python 🐍

The latest chapter of Joscelin Rocha's Learning-R-Resources repo will help you learn the basics of Python if you already know R. Also, check out the index on the side. There's a lot in this repo for R learners.
GitHub | Joscelin Rocha

Data Visualization

Science of Visual Data Communication: What Works.

Epic review of research-backed guidelines to determine what works and what doesn't work in visualizations. Includes lots of examples, linked references, and a summary of key guidelines. This is essential reading for leveling-up your data visualization proficiency.
Sage Journals

Python power-up: new tool visualizes complex data

Napari is an interactive, image viewer for Python that's designed for browsing, annotating, and analyzing large multi-dimensional images. It's important because Python has become a key language for scientific computing but it doesn't handle arbitrarily complex data. This article in Nature is a nice overview of Napari and the problems it solves.
Nature | Jeffrey M. Perkel

How did you like this issue of Data Elixir?

👎    1     2     3     4     5   👍

Was this email forwarded to you?  Sign up here >>