Rather than using neural nets as black boxes, they use ML intrepretability tools to get new insights into nonlinear distortions in optical signals, and design signal processing methods to improve optical communications.
— hardmaru (@hardmaru) September 11, 2020
paper https://t.co/dDiMZ9UNMF
blog https://t.co/ACOOzvgXl0 pic.twitter.com/ysnMdtCa7l
Their paper “Transferring Inductive Biases through Knowledge Distillation” (https://t.co/EePbSjaVjw) investigating inductive biases from knowledge distillation process (w/ @m__dehghani @wzuidema)
— hardmaru (@hardmaru) September 8, 2020
Also check out related blog postshttps://t.co/7AYyNJ2MHwhttps://t.co/3wMXjNQYWK
It's Hard for Neural Networks To Learn the Game of Life; new paper that examines how some “neural networks rely on lucky random initial weights of subnetworks called “lottery tickets” that converge quickly to a solution”. https://t.co/zGU893S6IJ
— Gary Marcus (@GaryMarcus) September 7, 2020
Thread: Last week, a list of 100 important NLP papers (https://t.co/PUHTvKCuiI) went viral. The list is okay, but it has almost *no* papers with female first authors.
— Sasha Rush (@srush_nlp) September 7, 2020
NLP is rich with amazing female researchers and mentors. Here is one paper I like for each area on the list:
Great application of #tensors Multi head attention is redundant and #tensor decomposition helps with compression. Great to see #tensorly used @JeanKossaifi @NVIDIAAI https://t.co/TQSX8osnlG
— Prof. Anima Anandkumar (@AnimaAnandkumar) September 7, 2020
TiVGAN: Text to Image to Video Generation with Step-by-Step Evolutionary Generator
— AK (@ak92501) September 7, 2020
pdf: https://t.co/jS8ZQDsk0x
abs: https://t.co/5BYGvtGJOj pic.twitter.com/KdjCmHtauE
Wonderful to see some theory behind the great success of self-supervised learning. Still trying to get our slow brains around how strong the results are. Cameo for the Stanford Sentiment Treebank—can it become the MNIST of #NLProc? By @jasondeanlee & al. https://t.co/BBmPDAH1Yk pic.twitter.com/TZd6vL3E7K
— Stanford NLP Group (@stanfordnlp) September 6, 2020
RAFT: Recurrent All Pairs Field Transforms for Optical Flow https://t.co/l7dfHUycS4 ECCV 2020 best paper award, fun reading. Like the careful use of inductive biases in the architecture design, other small details (e.g. convex combo upsampling). Shocked to see no Transformers :p pic.twitter.com/ubInOPe8cH
— Andrej Karpathy (@karpathy) September 6, 2020
A recent AutoML eval.: "Can AutoML outperform humans? An evaluation on popular OpenML datasets using AutoML Benchmark" Answer: Yes, sometimes, and if not, AutoML is not that far off. Good news, I guess; AutoML is not as bad as most people think it is. https://t.co/v3Xp3YaGTg pic.twitter.com/7o27kNlIZM
— Sebastian Raschka (@rasbt) September 4, 2020
We've used reinforcement learning from human feedback to train language models for summarization. The resulting models produce better summaries than 10x larger models trained only with supervised learning: https://t.co/Sk31d1CnTu
— OpenAI (@OpenAI) September 4, 2020
A Wholistic View of Continual Learning: https://t.co/DnOrnfpAYu
— Denny Britz (@dennybritz) September 4, 2020
If you're interested in Continual Learning and Active Learning, this paper comes with a neat taxonomy! pic.twitter.com/WJYWoynMvQ
Super excited to share our work on video completion at #ECCV2020!🤩Our method seamlessly removes objects, watermarks, or expands field-of-view from casually captured videos.
— Jia-Bin Huang (@jbhuang0604) September 4, 2020
Paper: https://t.co/JwT9bYN85V
Project: https://t.co/7raLq7jVQx
With @gaochen315, Ayush, and @JPKopf pic.twitter.com/eLGjINlbMC
