Sean Koenig aka DataCascadia

University of Washington Data Science TA - tweets mostly about 🐍 #python 📊 #datascience 📈 #datavis ; IRL enthusiastic about 🐕🌲🗺️✊

[Sean Koenig aka DataCascadia @datacascadia]

@yohaniddawela @threadreaderapp unroll

[Sean Koenig aka DataCascadia @datacascadia]

@nomic_ai @srush_nlp @llama @MongoDB @threadreaderapp unroll

[Sean Koenig aka DataCascadia @datacascadia]

@aparnadhinak @OpenAI @ArizePhoenix @arizeai @threadreaderapp unroll

[Aditya Kusupati @adityakusupati]

Introducing💃AdANNS: A Framework for Adaptive Semantic Search🕺

TL;DR: Up to 90× faster nearest neighbor retrieval and 2× lower memory cost for web-scale search.

Applies to vector search at scale & improves all "retrieval" augmented models!

https://t.co/yEnrr4oCwa

[1/8]

[Qiusheng Wu @giswqs]

The Open Geospatial Software (opengeos) GitHub organization is up and running! I have transferred 23 open-source #geospatial projects from my personal GitHub account to this org. We always welcome new contributors and collaborators. 👇
GitHub: https://t.co/pUJ5bSuz9f

[Cole Knaflic @storywithdata]

Two questions for you...

1. Can I put together inspiring content for an informative & interactive virtual *storytelling with data* mini-workshop that is streamed live & open/free to everyone around the world by next week?

2. If I were to do so, would you attend?

[Jeremy Howard @jeremyphoward]

There's a lot of folks under the misunderstanding that it's now possible to run a 30B param LLM in <6GB, based on this GitHub discussion.

This is not the case. Understanding why gives us a chance to learn a lot of interesting stuff! 🧵
https://t.co/9xOfNyLUFr

[Sumanth @Sumanth_077]

"Jax" is a python library for numerical computing, machine learning and is really faster than Numpy 🚀

Here is why you should care about Jax and its really powerful features in Machine Learning:🧵

[Cameron R. Wolfe, Ph.D. @cwolferesearch]

Is training deep neural nets from scratch about to become obsolete? Recent research indicates that the answer might be yes. Here's why...

📰: https://t.co/KRWOXcoxC9

[1/5]

[Aran Komatsuzaki @arankomatsuzaki]

Learning Performance-Improving Code Edits

Allows CODEGEN to match the performance of 10x bigger CODEX on the task of suggesing functionally correct, performance improving code edits.

proj: https://t.co/aYNOWIEqFd
abs: https://t.co/CzarPSXKGX

[Sean Koenig aka DataCascadia @datacascadia]

Fantastic long read taking you from what is a neural net to the implications for semantic logic of ChatGPT 🤯

https://t.co/BjaPYBUZNi

[Nat Henry 🦋 @nathenry.com @NatMakesMaps]

I made an interactive map showing every single block in Seattle where you can live within walking distance of daily necessities:
https://t.co/2cs5RDF4kR

[Alex Gutman @GutmanDataHead]

Prediction intervals for machine learning models.

MAPIE works with any scikit-learn model & is a big improvement over basic methods that only gave constant interval widths. Based on conformal prediction. Awesome library. https://t.co/NGZJaA79I6

[Stefano Giomo @stgiomo]

The island of real images is a metaphor to help you visualize how Stable Diffusion works.
Let the U-NET guide you in this journey from the sea of noise to the peak of sharpness at https://t.co/ztx4yOpAZJ

[Santiago @svpino]

11 ways ChatGPT saves me hours of work every day, and why you'll never outcompete those who use AI effectively.

A list for those who write code:

1 of 16

[Jordan Tigani @jrdntgn]

I wrote a blog post about the end of big data https://t.co/3H8l2SFhrg

[amy @a_e_roberts]

Processing images in transformers just became 100x better.

Since 4.25, we've been working on a new class: ImageProcessor.

Here's everything you need to know 👇 🧵 1/

[Cameron R. Wolfe, Ph.D. @cwolferesearch]

Neural Radiance Fields (NeRFs) are one of the most notable deep learning-based approaches that can learn and generate photorealistic renderings of 3D environments. Here are the 5 main ideas/components that make NeRFs so effective... 🧵 [1/10]

[Sebastian Raschka @rasbt]

Once in a while, I stumble upon some nice little feature in the @PyTorch library that I haven't seen before.

Have you tried AutoAugment, yet?

Just gave it a try: takes a few epochs, but that's a 10% accuracy improvement in this simple example case
https://t.co/yo3LUeFbf1

[Pete @nonmayorpete]

Google's new music model MusicLM is the breakthrough of the week.

Here it is in action.

Just describe the music and it'll generate the track: