ISSUE 493 ·   July 16, 2024

Interviews

Mustafa Suleyman on Defining Intelligence

In 2010, Mustafa Suleyman co-founded and led the application of AI at DeepMind. Later, he became the CEO and founder of Inflection AI, a venture partner at Greylock in 2022, and today, he's leading Microsoft's AI efforts. This is a great discussion about the current state of AI, LLMs, and where things are going.
YouTube | with Seth Rosenberg - 34 minutes

PRESENTED BY Statsig

Statsig’s Experimentation Week: We’re Shipping a New Feature Every Day.

Statsig is the platform trusted by OpenAI, Anthropic, Notion and Brex to ship, test, and analyze every feature they launch. Automate experiment analysis, rollout features, and more.

We’re always building, and this week, we’re taking it to another level. It’s experimentation week: every day we’re launching a new feature to make experimentation easier and more reliable. Tuesday is Stratified Sampling — follow along on LinkedIn to see the feature and what we’ll launch tomorrow.

Follow along on LinkedIn

Posts & Tutorials

Starting an Analytics Org From Scratch

DoorDash VP Jessica Lachs has grown the company’s analytics team from a band of scrappy generalists to a highly specialized 300-person organization. In this post, she spells out how to lay the groundwork for an analytics team and how to pick the right first hires.
First Round Review | Jessica Lachs

ML Code Challenges

These code challenges are a great way to learn about machine learning. For each challenge, you can choose to read through a short tutorial on how to answer it, submit a solution and have the solution verified. 
Deep-ML

Use-cases for inverted PCA

A lot of people know of PCA for it's ability to reduce the dimensionality of a dataset. It can turn a wide dataset into a thin one, with only a limited amount of information. But what about doing it the other way around as well? Can you turn the thin representation into a wide one again? And if so, what might be a use-case for that?
YouTube | :probabl.

Annotated area charts with plotnine

The plotnine visualization library is based on ggplot2 and brings the Grammar of Graphics to Python. It's a powerful approach that gives you the ability to compose plots by mapping variables in a dataframe to the visual objects that make up the plot. This step-by-step tutorial uses a simple, annotated area chart to show how to use plotnine.
Nicola Rennie

⚽ Euro 2024 Data

To help support the development of the next generation of sports analysts, StatsBomb just released all of its tactical event data from all 51 matches of the recent Euro 2024 tournament. This post describes what the data is, how to access it, and code to make it manageable. If you're interested in sports analytics, this is a goldmine of data
StatsBomb

Tools & Code

• PyPI Scout helps you find packages on PyPI using natural language queries - Florian Maas
   
• nanoparquet is a new R package that reads Parquet files into data frames, and writes data frames to Parquet files - Tidyverse Blog
   
• TimeGPT-1 is the first foundation model for forecasting and anomaly detection - GitHub | Nixtla
   
• Inspectus provides visualization tools for machine learning that runs in Jupyter notebooks via a simple Python API - GitHub | labml.ai

Papers

Encoding Spreadsheets for Large Language Models

SpreadsheetLLM is designed to give LLMs the ability to process, analyze, and understand spreadsheets. The research includes an encoding/compression framework to help work within token constraints as well as a "Chain of Spreadsheet" framework for spreadsheet understanding and validation. This line of research unlocks a lot of possibilities.
arXiv | Microsoft Research

Last Issue's Top Links

• The Amazon Weekly Business Review - Cedric Chin
   
• The Orange Book of Machine Learning - Carl McBride Ellis
   
• Fun with Positron - Andrew Heiss

How did you like this issue of Data Elixir?

👎    1     2     3     4     5   👍

Was this email forwarded to you?  Sign up here >