Data Science #37

In this issue: data flywheels for LLM applications; density and likelihood: what’s the difference?; what happened to BERT & T5; responsible forecasting; the explosion in time series forecasting packages; annotated area charts with plotnine; and more.

The sponsor of this issue is GitNotebooks.

• Data flywheels for LLM applications
• Density and likelihood: What’s the difference?
• What happened to BERT & T5? On transformer encoders, PrefixLM and denoising objectives
• [Sponsored] Jupyter Notebook Review Tool for GitHub

Frustrated with GitHub’s notebook review experience? Find out how Data Science teams are cutting their notebook review time in half with GitNotebooks. Quickly share feedback while staying in sync with GitHub. (Did we mention it’s free for most teams?)

• [ArXiv] Responsible forecasting: Identifying and typifying forecasting harms
• Kernel trick - deep convolutional representations in Reproducing Kernel Hilbert Space
• How to extract analytics from Bluesky, the new open social network
• [NEW BOOK] Start reading Andriy's new book today. New chapters are coming soon!
• The explosion in time series forecasting packages
• [Tutorial] Annotated area charts with plotnine

More than 60,000 subscribers are reading this newsletter. If you build a data product or provide a data service, you can become a sponsor of one of the future newsletter issues and get your business featured in the newsletter. Feel free to reach out to [email protected] for more details on sponsorships.

Enjoy the newsletter? Please help us make it bigger and better by sharing it with colleagues and friends.