How to Snap Out of a Data Science Slump
Our weekly selection of must-read Editors’ Picks and original features
What do you do when things aren’t going the way you were hoping they would? Whether it’s a machine learning model that required too much tinkering, a job offer that never materialized, or just a passing “wait, it’s almost June?!” moment of terror, we all face a setback (or worse) every once in a while. We found inspiration in our recent conversation with data scientist and TDS author Carolina Bento, whose pragmatic approach to problem-solving might resonate with you, too: “Sometimes, when I get stuck, I think about how I can approach the problem from a different angle.”
Carolina’s advice goes into much greater detail, from breaking a problem into its smallest atoms to talking to a trusted peer (read it—you won’t regret it!), but the general idea applies across so many difficult situations in our day-to-day experience. Reframing our perspective can be difficult, but it often pays off—whether in a new solution, a deeper understanding, or an unexpected opportunity.
If you don’t believe us, read Mike Bostock’s post about the promise of JavaScript for the future of data analysis. After years of never-ending debate on the relative merits and shortcomings of Python, R, and Julia, Mike wonders if it might be the right time for a big, splashy change and to turn to the most common programming language among developers.
Along similar lines, Daniel Marcous proposes new ways to scale up data operations as companies grow. He’s synthesized them in the detailed account of his Full Cycle Data Science framework, and also discusses how he’s implemented the latter in his own work. Another area rife for a perspective shift? For our latest TDS Podcast guest, Eliano Marques, it’s AI privacy. Eliano chatted with host Jeremie Harris about the urgent need to educate end users about the tradeoffs of giving our data away and to focus on the safety of algorithms.
Change doesn’t have to be big and splashy, of course. Cumulative iteration and experimentation can be just as powerful; Hannah Wnendt’s latest deep dive is a case in point, following her adventures in code benchmarking and her eventual success at boosting computational speed for data analysis in R. Rebecca Vickery, meanwhile, points out that sometimes, it’s not our processes but our tools that hold us back—so perhaps it’s time to explore new options. For Rebecca, this involves complementing the ever-popular Jupyter Notebook with four additional tools that help workflows stay smooth and efficient.
For many of us, snapping out of a slump requires a nudge from the outside—something to rid our mind of its murky thoughts. What better way to achieve that than by learning something new? If that’s your go-to method, too, you’ll love our latest selection of hands-on tutorials and guides.
- Start with the latest crop of plain-English tutorials from Ketan Doshi, who shared no fewer than three new updates to his ongoing series. Explore image-caption generation with Attention in TensorFlow, move on to learn about two essential NLP metrics (Bleu score and WER), or, if you’re having a deep learning kind of day, head to Ketan’s visual explanation of Batch Norm.
- From the cutting-edge of supervised computer vision research, Klemen Kotar reflects on some of the challenges the field is facing, and how a new paper he co-authored aims to resolve them.
- Round off your weekly reading with Vasnetsov Andrey’s introductory tutorial on metric learning—including how to train metric learning models with no manually labeled data.
If your week has been fantastic so far, we hope it stays that way! If it hasn’t, we hope diving into some hearty data science reads helped. Either way, we’re grateful for being part of it, and for all you do to sustain this community—from sharing our work with your friends and colleagues to supporting us by becoming Medium members.
Until the next Variable,
TDS Editors
Recent additions to our curated topics:
Getting Started
- Deconstructing the Job-Hunting Process: A Personal Perspective by Jesse Ruiz (she/they)
- A Better Way to Visualize Decision Trees with the dtreeviz Library by Parul Pandey
- Generalized Linear Models Clearly Explained by Lily Chen
Hands-On Tutorials
- Transformers: Implementing NLP Models in 3 Lines of Code by Fernando López
- How to Predict Customer Churn from Your Website Logs by Suhong Kim
- Simulate Real-Life Events in Python Using SimPy by Khuyen Tran
Deep Dives
- Casual ML for Data Science:Deep Learning with Instrumental Variables by Haaya Naushan
- How to Automate 3D Point Cloud Segmentation and CLustering with Python by Florent Poux, Ph.D.
- Analytics Essentials for Data Science by Matt Sosna