Bayesian methods and what they offer compared to classical econometrics https://t.co/DGG64raX7d

— Andrew Gelman (@StatModeling) March 7, 2021

Here's the blog post about:

✨ Deploying FastAPI (and other) apps with HTTPS powered by @traefik ✨

It has all the additional explanations, config files, and source code. 🎉

And there's a link to the recorded video event too. 📺https://t.co/BBIaHh9gll

— Sebastián Ramírez (@tiangolo) March 6, 2021

Algorithms are not good at predicting what ads I want to see. So why do we keep using them to “predict” who will drop out of school, commit a future crime or be unable to pay their rent.

My newsletter this week:https://t.co/f4arZL78Bh

— Julia Angwin (@JuliaAngwin) March 6, 2021

Sudden drops in life expectancy across the world and their causes, 1850 -2020.

Does not include data from 2020 / #COVID19 yet. #dataviz

Source: https://t.co/H3SyC6Ronv pic.twitter.com/D7EOZbGk2u

— Randy Olson (@randal_olson) March 5, 2021

RAWGraphs got a 2.0 update. Same open-source tool, with improved drag-and-drop visualization https://t.co/q2cl8FQy6F pic.twitter.com/sedM9mdnLi

— Nathan Yau (@flowingdata) March 5, 2021

Map of the North American Power Grid includes the contiguous United States and the southern parts of Canada (most Canadians hug the US border so connecting grids makes sense). Source: https://t.co/QwKbQCsXnU pic.twitter.com/yQxvdIgxbB

— Simon Kuestenmacher (@simongerman600) March 5, 2021

One of my longest standing dreams for neural network interpretability is that it will expand the space of possible human thought (https://t.co/Ncd2qbCMn5). pic.twitter.com/IVyHG08cyI

— Chris Olah (@ch402) March 4, 2021

Detectron2Go (D2Go) is a new, state-of-the-art extension for Detectron2 that gives developers an end-to-end pipeline for training and deploying object detection models on mobile devices and hardware.https://t.co/SjtH6PWBQq pic.twitter.com/QeZE4rR74w

— Facebook AI (@facebookai) March 4, 2021

SEER: large-scale SSL for vision.
- pre-train via SSL on 1 billion randomly selected images using SwAV.
- fine-tune on ImageNet: 84.2% top-1 accuracy.
- ft on 10% of ImageNet: 77.9%
- ft on 1% (13 samples per class): 60.5%
- beats SOTA on other CV taskshttps://t.co/Q8BT5QvKmf

— Yann LeCun (@ylecun) March 4, 2021

A new blog post I wrote with Ishan Misra.
An overview of Self-Supervised Learning.

We look at recent progress in SSL for vision & explain why SSL is more challenging with high-D continuous signals (images, video) than it is for discrete signals (text).https://t.co/DlL885CPpb

— Yann LeCun (@ylecun) March 4, 2021

We keep updating BERTScore, our generation evaluation method, behind the scenes. Been a while so highlights:
- Now supports 53 pre-trained models via @huggingface's Transformers
- WMT-16 to-EN correlations here:https://t.co/nlcL2QxtGh --> current best: deberta-xlarge-mnli

— Yoav Artzi (@yoavartzi) March 3, 2021

I've seen a lot of people worry about being a slow programmer. in my experience, if you:

1. work on important projects
2. keep moving forward
3. finish the projects

it doesn't matter too much how fast or slow you are

— 🔎Julia Evans🔍 (@b0rk) March 3, 2021