Week Ending 3.20.2022
RESEARCH WATCH: 3.20.2022
SPONSORED BY
This week was active for "Computer Science", with 1,404 new papers.
The paper discussed most in the news over the past week was by a team at University of Illinois at Urbana-Champaign: "BaM: A Case for Enabling Fine-grain High Throughput GPU-Orchestrated Access to Storage" by Zaid Qureshi et al (Mar 2022), which was referenced 23 times, including in the article Nvidia wants to speed up data transfer by connecting data center GPUs to SSDs in ArsTechnica. The paper got social media traction with 8 shares. A Twitter user, @granderojo, commented "ahead on a project i think i'll do some light reading tomorrow".
Leading researcher Oriol Vinyals (DeepMind) published "Integrating Language Guidance into Vision-based Deep Metric Learning" @summarizedml tweeted "A language guidance objective for DML, a method for contextualizing and realigning visual representation spaces for better semantic consistency. 📄".
The paper shared the most on social media this week is by a team at University of Tromsø: "The Mathematics of Artificial Intelligence" by Gitta Kutyniok (Mar 2022) with 356 shares. @omarsar0 (elvis) tweeted "The Mathematics of Artificial Intelligence Nice article summarising the importance of mathematics in deep learning research and how it’s helping to advance the field. Great read for ML students".
The most influential Twitter user discussing papers is AK who shared "Surrogate Gap Minimization Improves Sharpness-Aware Training" by Juntang Zhuang et al (Mar 2022) and said: "Surrogate Gap Minimization Improves Sharpness-Aware Training abs: Empirically, GSAM consistently improves generalization (e.g., +3.2% over SAM and +5.4% over AdamW on ImageNet top-1 accuracy for ViT-B/32)".
This week was very active for "Computer Science - Artificial Intelligence", with 228 new papers.
The paper discussed most in the news over the past week was "Compute Trends Across Three Eras of Machine Learning" by Jaime Sevilla et al (Feb 2022), which was referenced 12 times, including in the article Why changing computing trends across different ML eras matter in Analytics India Magazine. The paper author, Tamay Besiroglu, was quoted saying "Seeing so many prominent machine learning folks ridiculing this idea is disappointing". The paper also got the most social media traction with 494 shares. The investigators study trends in the most readily quantified factor - compute. A Twitter user, @TShevlane, posted "Remember the year 2010? We now have AI systems that take roughly 10 billion times more compute to train than back then. Seems like an important shift!", while @ohlennart observed "Compared to AI and Compute we find a slower, but still tremendous, doubling rate of 6 months instead of their 3.4 months. We analyze this difference in Appendix E 5/".
Leading researcher Pieter Abbeel (UC Berkeley) came out with "SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning".
The paper shared the most on social media this week is "Efficient Language Modeling with Sparse all-MLP" by Ping Yu et al (Mar 2022) with 231 shares. The investigators analyze the limitations of MLPs in expressiveness, and propose sparsely activated MLPs with mixture - of - experts (MoEs) in both feature and input (token) dimensions. @kawamuramasahar (川村正春 @ 五城目人工知能アカデミー) tweeted "propose sparsely activated MLPs with mixture-of-experts (MoEs) in both feature and input dimensions improves language modeling perplexity and obtains up to 2x improvement in training efficiency compared to Transformer-based MoEs and dense Transformers".
The most influential Twitter user discussing papers is Jordan Ellenberg who shared "Bertrands Postulate for Carmichael Numbers" by Daniel Larsen (Nov 2021) and said: "Here's Larsen's paper on Carmichael numbers, which finished 4th in this year's Regeneron".
This week was very active for "Computer Science - Computer Vision and Pattern Recognition", with 415 new papers.
The paper discussed most in the news over the past week was by a team at Google: "Kubric: A scalable dataset generator" by Klaus Greff (Derek) et al (Mar 2022), which was referenced 5 times, including in the article Googlers and co offer video dataset-generating Kubric in TheRegister.com. The paper got social media traction with 422 shares. A Twitter user, @giffmana, posted "This is one of our CVPR papers. If none of the coauthors write a summary thread, I'll do so later on. I contributed the small pretrain-transfer part. This giant collaboration was led by Klaus Greff and ; Klaus did the lion's share of work".
Leading researcher Oriol Vinyals (DeepMind) published "Non-isotropy Regularization for Proxy-based Deep Metric Learning".
The paper shared the most on social media this week is by a team at University of Maryland: "Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility and Double Descent from the Decision Boundary Perspective" by Gowthami Somepalli et al (Mar 2022) with 151 shares. @ducha_aiki (Dmytro Mishkin 🇺🇦) tweeted "Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility... Liam Fowl, Arpit Bansal tl;dr: decision-boundary smoothness depends on NN width, optimizer and label noise".
The most influential Twitter user discussing papers is Jordan Ellenberg who shared "Bertrands Postulate for Carmichael Numbers" by Daniel Larsen (Nov 2021)
Over the past week, 26 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was "Compute Trends Across Three Eras of Machine Learning" by Jaime Sevilla et al (Feb 2022)
The paper shared the most on social media this week is by a team at Harvard University: "Tweets in Time of Conflict: A Public Dataset Tracking the Twitter Discourse on the War Between Ukraine and Russia" by Emily Chen et al (Mar 2022) with 53 shares. @loriweiss310 (lori weiss) tweeted "& are "doing their part to enable the research community to understand Twitter chatter about the #UkraineWar " Publicly sharing dataset of 63M tweets. Social media in times of conflict".
The most influential Twitter user discussing papers is Jay Van Bavel who shared "Moral Emotions Shape the Virality of COVID-19 Misinformation on Social Media" by Kirill Solovev et al (Feb 2022) and said: "Moral Emotions Shape the Virality of #COVID19 Misinformation An analysis of 10,610 rumor cascades retweeted >24 million times finds that COVID-19 misinformation is more likely to go viral than facts because it contains more other-condemning emotions".
This week was very active for "Computer Science - Human-Computer Interaction", with 36 new papers.
The paper discussed most in the news over the past week was "Open5x: Accessible 5-axis 3D printing and conformal slicing" by Freddie Hong et al (Feb 2022), which was referenced 2 times, including in the article Affordable 5-Axis 3D Printer Upgrade Concept in Fabbaloo. The paper got social media traction with 6 shares.
This week was extremely active for "Computer Science - Learning", with 477 new papers.
The paper discussed most in the news over the past week was by a team at Curtin University: "Successful Recovery of an Observed Meteorite Fall Using Drones and Machine Learning" by Seamus L. Anderson et al (Mar 2022), which was referenced 20 times, including in the article For First Time, A Fresh Meteorite's Exact Location Has Been Found Using A Drone in IFLScience. The paper author, Seamus Anderson, was quoted saying "A camera-fitted drone flies over and collects images of the fall zone, which are transferred to our field computer where an algorithm scans each image for meteorites and features that resemble them". The paper got social media traction with 90 shares. A Twitter user, @PaperTldr, said "🗜91% We present the first time recovery of a fresh meteorite using a drone learning algorithm", while @summarizedml posted "We report the first-time recovery of a fresh meteorite fall using a drone and a machine learning algorithm. 📄".
Leading researcher Pieter Abbeel (UC Berkeley) came out with "SURF: Semi-supervised Reward Learning with Data Augmentation for Feedback-efficient Preference-based Reinforcement Learning".
The paper shared the most on social media this week is by a team at University of Tromsø: "The Mathematics of Artificial Intelligence" by Gitta Kutyniok (Mar 2022)
The most influential Twitter user discussing papers is Jay Van Bavel who shared "Moral Emotions Shape the Virality of COVID-19 Misinformation on Social Media" by Kirill Solovev et al (Feb 2022)
This week was active for "Computer Science - Multiagent Systems", with 26 new papers.
Over the past week, 18 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 Google: "Block-Recurrent Transformers" by DeLesley Hutchins et al (Mar 2022), which was referenced 1 time, including in the article Google & IDSIA’s Block-Recurrent Transformer Dramatically Outperforms Transformers Over Very Long Sequences in SyncedReview.com. The paper got social media traction with 209 shares. On Twitter, @Yuhu_ai_ observed "How do you make a transformer recurrent? You just turn the transformer 90 degree, and apply it in the lateral direction! Now, with recurrence, the context size is infinite! Let's make the recurrence great again with Block-Recurrent Transformers".
This week was very active for "Computer Science - Robotics", with 104 new papers.
The paper discussed most in the news over the past week was "Robotic Telekinesis: Learning a Robotic Hand Imitator by Watching Humans on Youtube" by Aravind Sivakumar et al (Feb 2022), which was referenced 3 times, including in the article Robotic telekinesis: Allowing humans to remotely operate and train robotic hands in Tech Xplore. The paper author, Deepak Pathak (Carnegie Mellon University), was quoted saying "Prior works in this area rely either on gloves, motion markers or a calibrated multi-camera setup". The paper got social media traction with 16 shares. A user, @summarizedml, tweeted "We build a system that enables any human to control a robot hand and arm, simply by demonstrating motions with their own hand. 📄", while @TheRealDAGsHub said "Robotic Telekinesis: Controlling Multifingered Robotic Hand by Watching Humans on Youtube. Paper: Project website".
Leading researcher Abhinav Gupta (Carnegie Mellon University) published "RB2: Robotic Manipulation Benchmarking with a Twist" The authors re - imagine benchmarking for robotic manipulation as state - of - the - art algorithmic implementations, alongside the usual set of tasks and experimental protocols.