If you have to watch *one* video from NeurIPS, it should be the keynote of Charles Isbell
— Thomas Wolf (@Thom_Wolf) December 9, 2020
It's not a keynote, it's a full movie (with special effects!) on AI, NLP, CV, bias and the broader & wider view on what we do and why we do it!
It's public here 👇https://t.co/CKkNeZmy38 https://t.co/Lyz02p8Wfc
Naturally Occurring Equivariance in Neural Networks -- A new Distill article by @ch402, @nickcammarata, @csvoss, @ludwigschubert, and @gabeeegoooh. This is the fourth article in the circuits thread.https://t.co/eeBuNI2D8g
— Distill (@distillpub) December 8, 2020
Well-written blog post on doing neuroevolution with Keras and PyGAD: https://t.co/q5Uxeo8Lgm
— François Chollet (@fchollet) December 8, 2020
Love Figure 8. of https://t.co/sJc3fcqHQe
— hardmaru (@hardmaru) December 8, 2020
“Images generated from different combinations of ingredient list and style noise. Images in the same row are generated with identical style noise.” https://t.co/Svx46eNmas pic.twitter.com/AnOjtuD3Sx
You can watch my talk from today's NeurIPS tutorial on abstraction & reasoning here: https://t.co/363V5ZM6gJ
— François Chollet (@fchollet) December 8, 2020
I've entirely retyped the caption file, so please turn on the subtitles and select "English"
Excited to share my latest blog which focuses on why recommender systems aren't faster on GPU and how you can fix that. TLDR; RecSys models are different from NLP and CV and require specialized dataloaders to get data to the GPU efficiently. https://t.co/Luey8PNy0E
— Even Oldridge (@Even_Oldridge) December 7, 2020
We added the visualisation of the attention patterns of Vision Transformer!https://t.co/zKZGRLCTmp
— Jean-Baptiste Cordonnier (@jb_cordonnier) December 7, 2020
Some heads learn translation equivarient attention to extract patches at fixed shifts. Other heads rely on color similarity or maybe more semantic features deeper in the network. pic.twitter.com/fHIXRQR1fS
The other day I trained a @PyTorch model in Oxford using data in Australia - without moving the data out of Australia.
— Andrew Trask (@iamtrask) December 6, 2020
Video: https://t.co/B4VjwLWQnb
It's a pretty cool tool called "Duet" - check it out.
It's probably going to unlock a lot of new data for you#100DaysOfMLCode
This chart shows how many leaders the Queen saw coming and going. While this is simply meant to show just how long the Queen has reigned for all I can think of that it's crazy to have unelected officials in the first place. Source: https://t.co/5hFWXIOJOz pic.twitter.com/6rfv7eQLXM
— Simon Kuestenmacher (@simongerman600) December 6, 2020
One thing the @timnitGebru situation highlights = fragility of having AI research center on a small number of private sector firms. If we substantially increased funding for AI research in academia (including large compute infrastructure), more research could take place there.
— Jack Clark (@jackclarkSF) December 5, 2020
NODE for Deep Learning on Tabular Data #ICLR2020
— ML Review (@ml_review) December 5, 2020
By @YandexAI
– DNN architecture that substantially outperforms leading GBDT on tabular data
– Generalizes CatBoost, making decision tree routing differentiable
ArXivhttps://t.co/CEgce5kXV6
PyTorchhttps://t.co/gq0hIkkf1y pic.twitter.com/vvX3F1cDMv
Our integration between TensorFlow Cloud and Kaggle notebooks has landed!
— François Chollet (@fchollet) December 4, 2020
Check it out: https://t.co/FL6VvmgKQ0
TensorFlow Cloud is an API for turning TF/Keras scripts & notebooks into distributed training jobs in the Cloud in one line. pic.twitter.com/smAk19cUAW
