Big data paradox. Data "scientists"...? ML playgrounds. Data serialisation in R. Building models like open-source software. ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌

ISSUE 366 · December 14, 2021

Insight

Why is “Data Scientist” such a controversial title?

There's a lot of debate about what "data science" is, exactly, and if "Data Scientists" should actually be called something else. The real problem is that "industry hasn’t learned how to take advantage of the special powers that 'scientists' bring to the table."
Emily Thompson

Big Surveys and the Big Data Paradox

Increasing the amount of data you collect may shrink confidence intervals but it will also magnify the effect of any survey bias. This paper uses recent vaccine surveys to show that happens. Ultimately, quality data is better than quantity data.
Nature

Tutorials, Projects & Opinions

A Call to Build Models Like We Build Open-Source Software

Pre-trained models are expensive to create and they're rarely updated, which means most of the research community gets left out of their design and creation. A lot could be gained by treating models more like open-source software. Here's how that could work.
Colin Raffel

Data serialisation in R

Great post that shows how R serializes an in-memory data structure to create a sequence of bytes that can be saved or transmitted.
Danielle Navarro

Analyzing scraped data using Git and SQLite

git-scraping makes it easy to record snapshots of an online data source to a repository for tracking changes over time. But then what? git-history takes that repository and converts it to a SQLite database so it can be easily worked with. Combine that with something like Datasette and you can do some pretty cool stuff.
Simon Willison

How to prep a new Apple machine for data science

Got a new Apple M1, M1 Pro or M1 Max? Here's a step-by-step guide to make sure you set it up correctly for TensorFlow and common data science libraries using Conda.
Daniel Bourke

How to make AI & BI work at scale

Join leading technologists from DataRobot, Buoy Health, and AtScale for a panel on how to unlock the power of data and analytics for everyone, foster more self-service and greater data science literacy, and generate a better return on data science investments.
Wednesday, December 15, 2021 at 2PM ET
// sponsored

Code & Tools

Introducing the 🤗 Data Measurements Tool

The 🤗 Data Measurements Tool is an open-source project for building, measuring, and comparing NLP datasets -- no programming required!
Hugging Face

Cogram

Cogram is a coding assistant for data science that's being built by a new YC startup. Cogram can generate Python code in Jupyter Notebook, and SQL from plain language, allowing data scientists to work more productively without looking up code syntax. It's available for free and they're especially looking for teams that might want customizations.
Cogram

Resources

Interactive Tools for ML, DL and Math

If you like interactive playgrounds, this is a great collection of learning tools for Deep Learning, Machine Learning, Interpretability and Math.
GitHub | Machine Learning Tokyo

Data Visualization

Plotly Resampler

Plotly is an awesome interactive visualization library but it gets slow when there are a lot of data points. To fix that, this library downsamples the data that are in a view and then plots the downsampled points.
GitHub | predict-idlab

Interview with Leland Wilkinson

Great interview with the late Leland Wilkinson where he describes highlights from his career and especially his book, The Grammar of Graphics. Leland was a legend in the data visualization community and sadly, he died at home last Friday.
PolicyViz, Episode #201

Outlier

Financial Modeling World Cup Tournament

Move over, League of Legends. The real future of esports is spreadsheets and Microsoft Excel!
PCWorld

How did you like this issue of Data Elixir?

👎 1 2 3 4 5 👍

To find specific content from prior issues or to research topics, check out the searchable Archives on Data Elixir's Search Page >>

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

Data Elixir - Data Elixir - Issue 366