World models are the future and the future is now! 🌎🚀
— Danijar Hafner (@danijarh) October 7, 2020
Proud to share DreamerV2, the first agent that achieves human-level Atari performance by learning behaviors purely within a separately trained world model.
Paper: https://t.co/deRLWwVZ8Q
Thread 👇 pic.twitter.com/KvCNKehBvp
I’m excited to announce our new Generative ENtity REtrieval (GENRE): the first system that retrieves entities by generating their unique names, autoregressively fully conditioned on context -> https://t.co/IVx4hdjKTq w/ @facebookai @AmsterdamNLP @ENS_ULM @ucl_nlp [1/4] pic.twitter.com/p4BHucSKJi
— Nicola De Cao (@nicola_decao) October 7, 2020
We've updated the EATS paper on arXiv: https://t.co/J8PgR8U2JG
— Sander Dieleman (@sedielem) October 6, 2020
'End-to-end' has many possible interpretations – Table 5 in the appendix (p. 21) describes some of the many ways in which the TTS pipeline has been factorised into stages in the literature, for easier comparison. https://t.co/I8AMnJoWfm pic.twitter.com/ku0K7EJ5QA
Are we making meaningful progress on machine learning for EHR data?
— Andrew Beam (@AndrewLBeam) October 6, 2020
New preprint with @DavidRBellamy and Leo Celi tries to answer this question through the lens of benchmarks and the answer is, unfortunately, "probably not"
Paper: https://t.co/KjgjvC0eWQ
Some highlights 👇 pic.twitter.com/5xZf5iqJQs
Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images
— hardmaru (@hardmaru) October 6, 2020
“Very Deep VAEs” achieve higher likelihoods, use fewer parameters, generate samples 1000x faster, and are more easily applied to hi-res images, compared to PixelCNN.https://t.co/bjjKrZvnb2 pic.twitter.com/u1xMHJiRpp
“Our models perform equivalently to backprop on ML benchmarks, while utilising only local and (mostly) Hebbian plasticity. Our method raises the potential that standard ML algorithms could in principle be directly implemented in neural circuitry.” 🧠https://t.co/jiZ2I1l73F pic.twitter.com/Pu4WflSqkr
— hardmaru (@hardmaru) October 5, 2020
🔥New Video🔥Convolutions are DEAD as Transformers continue to ruin absolutely everything 😱 New SotA on ImageNet, VTAB, etc using only Transformers + massive data 👑 Also Peer Review is broken. Watch Now!👀https://t.co/aQllZoKfdf@GoogleAI @giffmana @__kolesnikov__ @XiaohuaZhai pic.twitter.com/6jJf9FC3kw
— Yannic Kilcher (@ykilcher) October 4, 2020
GANs Can Play Lottery Tickets Too
— AK (@ak92501) October 3, 2020
pdf: https://t.co/8wo0UHsNrG
abs: https://t.co/WCaaZbgA97 pic.twitter.com/7xoX9lPhwQ
VQ-VAE + Transformer look much better than VAE + LSTMhttps://t.co/TBfxy3jysl https://t.co/NCfTl0eps5
— hardmaru (@hardmaru) October 3, 2020
Recent conversation with a friend:@ilyasut: what's your take on https://t.co/fqVhQNaBWQ?@OriolVinyalsML: my take is: farewell convolutions : ) pic.twitter.com/9PEvxmWvO4
— Oriol Vinyals (@OriolVinyalsML) October 3, 2020
An Attention Free Transformer
— AK (@ak92501) October 3, 2020
pdf: https://t.co/AH6w6QZeho
abs: https://t.co/IYkABa7rGw pic.twitter.com/AJgBMHQnWc
Image GANs meet Differentiable Rendering for Inverse Graphics and Interpretable 3D Neural Rendering
— AK (@ak92501) October 3, 2020
pdf: https://t.co/xP5VjzSOR5
abs: https://t.co/YDaxtE7spF pic.twitter.com/fdQvojpgyh
