Math Animations, Irreproducible Research, and Telling Stories with Data
“Story” is a word that sometimes feels overused—including in the context of data science. Not every slide deck with a clear structure and useful takeaways is a story, and that’s ok. But as Marie Lefevre argues in her post about compelling data storytelling, there are time-tested ways to make any analysis memorable and engaging, so why settle for dry and perfunctory?
TDS authors certainly don’t settle for dry and perfunctory; we’re frequently amazed by the range of narratives and voices we get to share. It takes great skill and effort to breathe life into topics that at first glance appear very technical. When you hear “reproducible research,” do you run straight to your kitchen to throw some popcorn into the microwave? Maybe you should, if the next item on your agenda is Vincent Vanhoucke’s call to recognize the value of irreproducible research, based on his robotics research work at Google.
While your popcorn is still fresh, hop over to Arnaud Guzman-Annès’s exploration of airline fleet optimization during the pandemic, a complex problem which he approaches with clarity and patience.
Do you still have a few kernels left? You won’t regret diving into Isabelle Augenstein’s study of Wikipedia entries of NLP scholars, which also discusses what these Wiki pages tell us about bias and representation in the field.
Teaching another person how to accomplish a concrete goal is a form of storytelling too, and the best teachers make their students forget they’re in the process of learning. Looking to sink your teeth into an engaging, fun tutorial? Start with Khuyen Tran’s guide, which shows how to create mathematical animations in Python, then proceed to Vishesh Khemani, Ph.D.’s fresh take on search algorithms, which pushes abstract concepts into the physical world (spoiler alert: books are involved).
Wrap up your learning session with TDS editor Elliot Gunn’s resource-filled feature on the importance of documentation—it includes some of the best articles we’ve published on this crucial-yet-often-neglected topic.
If your idea of a captivating story leans towards the speculative and the meta, have no fear: you won’t leave this newsletter empty-handed. You can spend the next 11 minutes contemplating Mark Saroufim’s thoughts on Lagrangian mechanics, drawing on the connections between machine learning and physics. Or grab your headphones, go for a leisurely stroll, and listen to Jeremie Harris’s conversation with Melanie Mitchell on the TDS Podcast, where they discuss the limits of AI and why a future in which we all serve our AI overlords might not be just around the corner.
Thank you for listening, reading, and engaging this week, and for opening up our community to so many conversations. And thank you for your support, too.
Until the next Variable,
TDS Editors
Recent additions to our curated topics:
Getting Started
- A Hands-On Demo of SQL vs. NoSQL Databases in Python by Matt Sosna
- A Checklist to Track Your Data Science Progress by Pascal Janetzky
- An Introduction to Bots by Sara A. Metwalli
Hands-On Tutorials
- Git Commands Data Scientists Use on a Day-to-Day Basis by Varshita Sher
- Visualizing Mexican Wildfires during 2017 with Folium by Isaac Arroyo
- Multiple Linear Regression Python 101 by Chuck Utterback
Deep Dives
- Neural Network Optimizers Made Simple: Core Algorithms and Why They Are Needed by Ketan Doshi
- Data Science in a World of Radical Uncertainty by James Lowe
- MLOps Best Practices for Data Scientists by Maggie MHANNA
