Check out AdvGen, a new approach to training neural machine translation models that applies generated adversarial examples to improve model robustness, reducing mistranslation rates and improving performance on standard benchmarks. Learn about it here! โ https://t.co/rzWIDSoVh1
โ Google AI (@GoogleAI) July 29, 2019
Relevant to this blog post I wrote a while ago: https://t.co/lyw14bLMZc
โ Rachael Tatman (@rctatman) July 29, 2019
Unfortunately it looks like even more evidence that common methods of ML algorithm evaluation have some pretty big flaws. ๐ข https://t.co/qgQ3DTY6FF
FixResNeXt is currently #1 in the Image Classification on ImageNet leaderboard! We propose a simple & efficient strategy to jointly optimize the train and test resolutions, which improves the classifier accuracy and/or reduces the training time. https://t.co/7n5uxUkNPz pic.twitter.com/9x5uxxs76F
โ Facebook AI (@facebookai) July 29, 2019
Introducing long-form question answering https://t.co/onDIQz6j5W
โ Yann LeCun (@ylecun) July 29, 2019
Our paper ๐๐ผ๐น๐ผ๐๐๐ก was accepted to #ICCV2019!
โ Thu Nguyen Phuoc (@thunguyenphuoc) July 29, 2019
We show that HoloGAN automatically learns a ๐ฑ๐ถ๐๐ฒ๐ป๐๐ฎ๐ป๐ด๐น๐ฒ๐ฑ 3D representation from natural images. NO pose labels, NO 3D shapes, NO multiple views, ONLY 2D images! https://t.co/p7zGmi1xYn
Video: https://t.co/J4xCYNLczZ pic.twitter.com/xtrKqy77x6
The Paper on RoBERTa, the current holder of the pole position on the GLUE leaderboard ( https://t.co/YBrziEPH2F )https://t.co/Dv2y0q83YA https://t.co/Dv2y0q83YA
โ Yann LeCun (@ylecun) July 29, 2019
An artificial neural network called โEmoNetโ can accurately recognize which emotions, out of 20 different categories, a human would feel in response to an image. Read more in the research from @ScienceAdvances: https://t.co/xrFzD2Dzsj #AI pic.twitter.com/Sywx7e8WpJ
โ Science Magazine (@sciencemagazine) July 28, 2019
Trends in Integration of Vision and Language Research:
โ ML Review (@ml_review) July 28, 2019
A Survey of Tasks, Datasets, and Methods [102pp]
By @nani852009 @dklakow
Covers 10 different tasks in terms of problem formulation, methods,
datasets, evaluation measures, SoTAhttps://t.co/edq4btL1JU pic.twitter.com/LWeGeIDVyk
Cool new network quantization/minimization method from FAIR-Paris and Universitรฉ de Rennes https://t.co/DEojJ1P0sV
โ Yann LeCun (@ylecun) July 26, 2019
See our original paper on PBT, which also forms the foundation of our StarCraft II program, AlphaStar, and our agents that reached human-level performance in Capture the Flag: https://t.co/X7hxjLWEen
โ DeepMind (@DeepMindAI) July 25, 2019
Good places for viewing โprogressโ in #NLProcโor at least the latest over-tuning results on various benchmarks๐: โhttps://t.co/4g5ScHzZGk โhttps://t.co/e9S1TOwdr0 โhttps://t.co/Ej92jujCPw โhttps://t.co/t0dNtQfpT9 โhttps://t.co/PkyAlQUf16 โhttps://t.co/tjaP7hT8ec
โ Stanford NLP Group (@stanfordnlp) July 24, 2019
New research shows how #machinelearning can improve high-performance computing for solving partial differential equations, with potential applications that range from modeling #climatechange to simulating fusion reactions. Learn all about it here โ https://t.co/Ei5m4CZe9W
โ Google AI (@GoogleAI) July 23, 2019
