Week Ending 3.6.2022
RESEARCH WATCH: 3.6.2022
SPONSORED BY
This week was active for "Computer Science", with 1,391 new papers.
The paper discussed most in the news over the past week was "Obstacle avoidance for blind people using a 3D camera and a haptic feedback sleeve" by Manuel Zahn et al (Jan 2022), which was referenced 31 times, including in the article This Device Lets Blind People See With Vibrations In Their Arms in Tech Register. The paper got social media traction with 8 shares. On Twitter, @stripp_lab posted "Quite some media coverage on this 'infrared vision' pre-print... Data not convincing though".
Leading researcher Yoshua Bengio (Université de Montréal) came out with "Combining Modular Skills in Multitask Learning" The authors assume that each task is associated with a subset of latent discrete skills from a (potentially small) inventory. @TLesort tweeted "(1/3) Performance decrease in continual learning is not only about forgetting! 😱🤯 A poor feature selection might be largely responsible for performance decrease. Learn more in: "Continual Feature Selection: Spurious Features in Continual Learning"".
The paper shared the most on social media this week is by a team at Microsoft: "DeepNet: Scaling Transformers to 1,000 Layers" by Hongyu Wang et al (Mar 2022) with 427 shares. The authors propose a simple yet effective method to stabilize extremely deep Transformers. @srchvrs (Leo Boytsov) tweeted ""Remarkably, on a multilingual benchmark with 7,482 translation directions, our 200-layer model with 3.2B parameters significantly outperforms the 48-layer state-of-the-art model with 12B parameters by 5 BLEU points, which indicates a promising scaling direction."".
The most influential Twitter user discussing papers is Francis Villatoro who shared "Integer versions of Yang-Mills theory" by R. A. Wilson (Feb 2022) and said: "Integer versions of Yang-Mills theory “In a recent paper, 't Hooft asks for an integer version of the gauge group of the standard model. Such groups were completely classified more than 100 years ago (the answer to 't Hooft's question).”".
This week was very active for "Computer Science - Artificial Intelligence", with 234 new papers.
The paper discussed most in the news over the past week was by a team at Microsoft: "Singularity: Planet-Scale, Preemptive and Elastic Scheduling of AI Workloads" by Dharma Shukla (Microsoft) et al (Feb 2022), which was referenced 22 times, including in the article Chip geopolitics — A call to ban predictive policing — Building the Metaverse in Politico.eu. The paper author, Mark Zuckerberg, was quoted saying "within our lifetimes." The paper got social media traction with 145 shares. On Twitter, @CKsTechNews observed "#Microsoft details 'planet-scale' AI infrastructure packing 100k-plus #GPUs Microsoft with the big mouth, less talk, more doing friends. Press Paper", while @PaperTldr observed "🗜87% Scheduling high utilization across deep learning and inference workloads is a crucial lever for cloud providers to train and deliver highly-efficient distributed service to their clients".
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) came out with "DIME: Fine-grained Interpretations of Multimodal Models via Disentangled Local Explanations".
The paper shared the most on social media this week is "FastFold: Reducing AlphaFold Training Time from 11 Days to 67 Hours" by Shenggan Cheng et al (Mar 2022) with 141 shares. The investigators propose FastFold, a highly efficient implementation of the protein structure prediction model for training and inference. @HochreiterSepp (Sepp Hochreiter) tweeted "ArXiv Speeding up AlphaFold: reduces training time from 11 days to 67 hours and 7.5∼9.5× speedup for long sequences. GPU optimizations. With github repository".
This week was very active for "Computer Science - Computer Vision and Pattern Recognition", with 367 new papers.
The paper discussed most in the news over the past week was by a team at UC Berkeley: "A ConvNet for the 2020s" by Zhuang Liu et al (Jan 2022), which was referenced 6 times, including in the article AI Papers to Read in 2022 in Towards Data Science. The paper got social media traction with 894 shares. On Twitter, @TacoCohen commented "👉 The first law of DL architectures 👈 "Whatever" is all you need 🤯 Any problem that can be solved by transformer / ViT can be solved by MLP / CNN, and vice versaSame for RNNs".
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) came out with "DIME: Fine-grained Interpretations of Multimodal Models via Disentangled Local Explanations".
The paper shared the most on social media this week is by a team at Tel Aviv University: "Generative Adversarial Networks" by Gilad Cohen et al (Mar 2022) with 142 shares. @summarizedml (SummarizedML) tweeted "Generative Adversarial Networks (GANs) are a generative adversarial network framework that achieve state-of-the-art imagegeneration 📄".
The most influential Twitter user discussing papers is Francis Villatoro who shared "Integer versions of Yang-Mills theory" by R. A. Wilson (Feb 2022)
Over the past week, 25 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was "AirGuard -- Protecting Android Users From Stalking Attacks By Apple Find My Devices" by Alexander Heinrich et al (Feb 2022), which was referenced 5 times, including in the article Experts Create Apple AirTag Clone That Can Bypass Anti-Tracking Measures in BusinessMayor.com. The paper got social media traction with 27 shares. A Twitter user, @maggied, posted "Have not dug into it yet but just saw this paper about AirGuard that they mention in the positive security post Also Re tracker detect is the AirTag paired or unpaired w Apple account? It worked for me with a paired one so long as out of range of owner".
The paper shared the most on social media this week is "Advancing an Interdisciplinary Science of Conversation: Insights from a Large Multimodal Corpus of Human Speech" by Andrew Reece et al (Mar 2022) with 202 shares. The researchers advance an interdisciplinary science of conversation, with findings from a large, novel, multimodal corpus of 1,656 recorded conversations in spoken English. @ProfAWBrooks (Alison Wood Brooks) tweeted "👏Major applause👏for this excellent scholarship and generative data sharing. Double congrats and thanks".
This week was very active for "Computer Science - Human-Computer Interaction", with 39 new papers.
The paper discussed most in the news over the past week was "Obstacle avoidance for blind people using a 3D camera and a haptic feedback sleeve" by Manuel Zahn et al (Jan 2022)
Leading researcher Sergey Levine (University of California, Berkeley) came out with "X2T: Training an X-to-Text Typing Interface with Online Learning from User Feedback".
This week was extremely active for "Computer Science - Learning", with 486 new papers.
The paper discussed most in the news over the past week was by a team at Microsoft: "DeepNet: Scaling Transformers to 1,000 Layers" by Hongyu Wang et al (Mar 2022)
Leading researcher Yoshua Bengio (Université de Montréal) came out with "Continuous-Time Meta-Learning with Forward Mode Differentiation" @summarizedml tweeted
This week was active for "Computer Science - Multiagent Systems", with 20 new papers.
The paper shared the most on social media this week is by a team at DeepMind: "Learning Robust Real-Time Cultural Transmission without Human Data" by Cultural General Intelligence Team et al (Mar 2022) with 91 shares. @summarizedml (SummarizedML) tweeted "A method for generating zero-shot, high recall cultural transmission in artificially intelligent agents. 📄".
Over the past week, 23 new papers were published in "Computer Science - Neural and Evolutionary Computing".
The paper discussed most in the news over the past week was by a team at Kyushu University: "Gen\eLive! Generating Rhythm Actions in Love Live!" by Atsushi Takada et al (Feb 2022), which was referenced 2 times, including in the article 你玩的音游可能真是AI生成的，Love Live!工作室发了篇论文，用AI节省50%作谱时间 in QbitAI.com. The paper got social media traction with 96 shares. On Twitter, @PaperTldr commented "🗜88% Rhythm action game is a music - based video game in which the player is challenged issue commands during a music session", while @fly51fly said "GenéLive! Generating Rhythm Actions in Love Live! A Takada, D Yamazaki, L Liu, Y Yoshida, N Ganbat, T Shimotomai, T Yamamoto, D Sakurai, N Hamada(2022) #MachineLearning #ML".
This week was extremely active for "Computer Science - Robotics", with 169 new papers.
The paper discussed most in the news over the past week was by a team at Microsoft: "Spatial Computing and Intuitive Interaction: Bringing Mixed Reality and Robotics Together" by Jeffrey Delmerico et al (Feb 2022), which was referenced 2 times, including in the article Researchers enhance human-robot interaction by merging mixed reality and robotics in Tech Xplore. The paper got social media traction with 17 shares. A Twitter user, @summarizedml, said "Spatial computing -- the ability of devices to be aware of their surroundings and to represent this digitally -- offers novel capabilities in human-robot 📄".