Metrics, Metrics, and More Metrics, Plus: Our On-Demand Beginners’ Guide Is Here
This edition of the Variable comes to you loaded, as always, with some of the best TDS reads of the past week. Before we get to them, though, we wanted to share a new resource we’ve just recently launched—our free, on-demand, email-based beginners’ guide.
If you’re currently taking your first few steps in data science, you can sign up to receive a daily dose of beginner-friendly articles and practical tips. (It’s a two-week curriculum, and you can unsubscribe at any time.) We hope you enjoy it.
Alright, reading recommendations! Let’s start with Joseph Rocca and Baptiste Rocca’s deep dive into the intricacies of the exploration-exploitation dilemma. They cover a tricky part of any data-informed decision-making process—the moment when you determine whether you have enough data for your needs, or should wait longer before the best course of action becomes apparent.
Continuing with the theme of fuzzy situations, Cassie Kozyrkov explores the thorny world of metrics, and insists that it’s crucial to stick to what you can define with precision: “instead of falling in love with a word and pursuing it for its own sake, think deeply about what real-world quantity you want to measure.”
If metrics can be messy, what can we say about AI ethics? In this fast-evolving field, progress seems within reach in theory, but things often get a lot murkier in practice. David Graus argues that we should apply fairness guidelines to real-world problems, and provides several powerful examples of what that might look like in areas as diverse as news recommendations and hiring. Still in the mood for a smart discussion of the future of AI? Don’t miss the latest TDS Podcast episode, where Jeremie Harris and guest Ryan Carey discuss safety, risk-management, and how we should design the right incentives for AIs.
Ismael Kherroubi Garcia goes in a similar direction in his exploration of the meaning of data in data science. It’s only when we approach data the right way—with the principles of epistemic humility and diversity in mind—that “analyses result in actionable knowledge that is relevant to the real world.”
If you’re feeling more hands-on than lofty this week (or maybe both hands-on and lofty?), we’ve got several options you should consider, too:
- Learn from Nina Hristozova about the ROUGE score and how this metric can help you evaluate the quality of abstractive text summarization.
- Follow Leigh Johnson’s journey as she builds Print Nanny, an AI/ML-powered product, from concept to prototype to soft launch.
- Read our Author Spotlight Q&A with Kurtis Pykes, who shares several tips and lessons from his path towards freelancing success as a data scientist.
- Master a new skill—Steve Attila Kopias published a new step-by-step tutorial, teaching us how to bring together gridded datasets and web-based maps.
We hope you enjoy this week’s selection—if any of the posts resonated with you in particular, consider leaving a comment for its author. (There are few things writers appreciate more than interacting with their readers!)
We’re as grateful as ever for your support, and for choosing to join us on your data science adventures.
Until the next Variable,
TDS Editors
Recent additions to our curated topics:
Getting Started
- Introduction to Datasette: Explore and Publish Your Data in One Line of Code by Khuyen Tran
- Kubeflow Pipelines: How to Build Your First Kubeflow Pipeline from Scratch by Fernando López
- How Women Can Restart Their Career in Data Analytics by Ruchi Deshpande
Hands-On Tutorials
- Advanced Forecasting Using Bayseian Diffusion Modeling by Fraser Lewis
- A Comprehensive Study of Mixed-Integer Programming with JuMP on Julia (Part 2) by Ouaguenouni Mohamed
- Data Preparation for Geospatial Analysis and ML with Laguerre-Voronoi in Python by Sunayana Ghosh
Deep Dives
- Survival Models for Histopathology by Heather Couture
- Causal Inference for Fun and Profit by Naim Kabir
- AI-Computed Breast Cancer Risk Map by jean-baptiste charraud
