ISSUE 401  ·   August 23, 2022

Trends

How to Build a GPT-3 for Science

A GPT-3 for science could help us train machines to answer questions and synthesize new ideas based on research. Imagine if a researcher could propose an experiment and an AI model responded with useful suggestions based on related work that's already been done. This post explores some of the possibilitites and what it will take to get there.
Future | Josh Nicholson

Data Elixir Partners

Interview Query

Interview Query is an adaptive learning platform that will help you develop your skills and prepare for your next interview. Questions come from actual interviews and cover a wide range of topics in data science, analytics, SQL, machine learning, and more. Choose your own learning path and get an inside look at specific companies, interviews, and roles. 

Tutorials, Projects & Opinions

Replacing Static Authentication Detections with Anomaly Based Detections

Multi-factor Authentication (MFA) is commonly used to prevent account takeover attacks but a variety of real-world issues make MFA unreliable. Here's a clever alternative that uses Isolation Forests to analyze authentication data for anomalies. Even if you don't have any interest in MFA, this is a short tutorial with an innovative approach.
Compass True North | Ryan Glynn

Data Product Canvas 

The Data Product Canvas is a practical framework to help you think through the development of a new data product. It will look familiar if you've worked with Lean methods for startup development but this is specifically tailored for data. Ultimately, the goal here is to help you avoid creating the right solution to the wrong problem.
Leandro Carvalho

Simplicity is an Advantage but Complexity Sells 

Complexity sells because it signals things like effort, mastery, and innovation. But simple things are easier to understand, easier to build, and cost less to use. This is a great post to help frame your thinking.
Eugene Yan

What's the big deal about embedded key-value databases?

You're probably already familiar with basic SQL databases like PostgreSQL and MySQL but more and more, people are talking about embedded key-value stores. Here's a high-level tour of what these new databases are and why they're important.
Phil Eaton

How to measure a subcontinent

These days, it's easy to take mapmaking for granted but in the days before GPS and lasers, making maps was super labor intensive. In this deep dive, Randy Au explores the craft of data collection and geodesy without modern tools. This is a great read with lots of linked references along the way.
Counting Stuff | Randy Au

Resources

Getting started with R and R-spatial

If you're interested in making maps using R, this collection of workshop materials from FOSS4G is a good place to start. It's organized as a book and covers a variety of topics, including spatial data handling, working with vector data, working with raster data, and mapmaking using the R-spatial ecosystem.
Nicolas Roelandt and Jakub Nowosad

Intro to Neural Networks and Backpropagation

Andrej Karpathy recently left his position as the Director of AI at Tesla and is teaching on the web again! Andrej is well-known for his super clear explanations and in this video tutorial, he offers a step by step introduction to backpropagation and neural network training. This assumes minimal math and basic knowledge of Python.
YouTube | Andrej Karpathy

How did you like this issue of Data Elixir?

👎    1     2     3     4     5   👍

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