ISSUE 436 · May 23, 2023

Trends

Google "We Have No Moat & Neither Does OpenAI"

Google and OpenAI have enormous resources for training big models but there are a lot of ways for smaller projects to thrive. For starters, data quality scales better than data size. This is a leaked internal Google document that's been out for two weeks already but it keeps showing up in my Inbox and it's a good read.
SemiAnalysis | Dylan Patel and Afzal Ahmad

Sponsored Link

Webinar | Is the Modern Data Stack Enterprise Ready?

How should enterprise companies think about the modern data stack today? Join Mode's next webinar with folks from Gartner, VMWare and Mode for a fireside chat on the topic. RSVP here.​​​​​​​

Tutorials & Opinions

Linear Algebra

Gilbert Strang, the legendary math teacher at MIT, retired this week after teaching for 61 years. His Linear Algebra course, in particular, is worthwhile and the lecture videos and course materials are still available online for free.
MIT Open Courseware | Gilbert Strang

The Basics of Python Packaging in Early 2023

This how-to guide walks through modern Python packaging standards and explains how to decide which tools are most appropriate for you.
DrivenData | Jay Qi

CLI & version control for ELT

Data teams can now fetch data from anywhere, send data anywhere, and transform data their way. Download open source Meltano.
// sponsored

Tools & Code

privateGPT

Here's a new python package that gives you the power of LLMs for interacting with private and/or proprietary documents. privateGPT is 100% private and no data leaves your environment. It has a permissive Apache 2.0 license and is built with LangChain, GPT4All and LlamaCpp.
GitHub | Iván Martínez

whyqd: simplicity, transparency, speed

whyqd is a python-based toolkit that makes it easy to create schema-to-schema "crosswalks" for restructuring messy data to conform to a standardized metadata schema. Here's how that's useful.
Whyqd

Julia 1.9 Highlights

Julia 1.9 introduces significant improvements in time-to-first-execution, memory management, and features for package authors. Key highlights include native code caching, package extensions, heap snapshot capabilities, and an interactive thread pool for task prioritization.
The Julia contributors

Resources

Tidy Finance with R

This comprehensive text brings together all the tools the authors wished they had at the beginning of their graduate studies in finance. Features chapters on data organization, empirical asset pricing, linear models, portfolio optimization, and more. Free to read online.
Christoph Scheuch, Stefan Voigt, Patrick Weiss

The World Is Built On Probability

This two-part book introduces probability from the perspective that all modern life is based on probability. This is a thought-provoking book that was originally written in 1984 and was recently updated and released for the web. Read online or follow the links to download.
MIR Publishers | Lev Tarasov

Data Visualization

Collection Space Navigator

The Collection Space Navigator is an interactive visualization tool for working with large image collections and their multidimensional representations. It provides customizable projections and a variety of filters that make it easy to get slices of high-dimensional space. There's a lot to explore here, including a paper, code, and a demo.
Tillmann Ohm, Mar Canet Sola et al.

How did you like this issue of Data Elixir?

👎    1     2     3     4     5   👍

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