Week Ending 12.5.2021
RESEARCH WATCH: 12.5.2021
This week was extremely active for "Computer Science - Artificial Intelligence", with 257 new papers.
The paper discussed most in the news over the past week was by a team at Washington University in St. Louis: "Learning to Compose Visual Relations" by Nan Liu et al (Nov 2021), which was referenced 11 times, including in the article Artificial Intelligence That Understands Object Relationships – Enabling Machines To Learn More Like Humans Do in SciTechDaily. The paper author, Yilun Du (Google), was quoted saying "When I look at a table, I can’t say that there is an object at XYZ location. Our minds don’t work like that. In our minds, when we understand a scene, we really understand it based on the relationships between the objects. We think that by building a system that can understand the relationships between objects, we could use that system to more effectively manipulate and change our environments". The paper got social media traction with 41 shares. On Twitter, @ThomasW423 said "Tenenbaum papers are the best".
Leading researcher Pieter Abbeel (University of California, Berkeley) came out with "Zero-Shot Text-Guided Object Generation with Dream Fields", which had 24 shares over the past 2 days. @PhilRead tweeted "Coming to an architect near you: “Computer: Design a modern home for a family of five based on an aesthetic sourced from all publicly available social media profiles of my client suitable for entertaining along the coast of North Carolina for a budget of $400/sf.”".
The paper shared the most on social media this week is by a team at University of Oxford: "The signature and cusp geometry of hyperbolic knots" by Alex Davies et al (Nov 2021) with 316 shares. @AnnUkhanova (Anna Ukhanova) tweeted "Amazing results from announced today! Fantastic example of what a deep collaboration with top domain experts can lead to! 🧑💻🤝🤖 Very excited about the impact that AI has on advancing science and congrats to everyone involved in this project! 👏".
The most influential Twitter user discussing papers is AK who shared "FQ-ViT: Fully Quantized Vision Transformer without Retraining" by Yang Lin et al (Nov 2021) and said: "FQ-ViT: Fully Quantized Vision Transformer without Retraining abs: 85.17% Top-1 accuracy with ViT-L on ImageNet and 51.4 mAP with Cascade Mask R-CNN (Swin-S) on COCO, achieve comparable accuracy degradation (∼1%) on fully quantized Vision Transformers".
This week was very active for "Computer Science - Computer Vision and Pattern Recognition", with 468 new papers.
The paper discussed most in the news over the past week was by a team at Washington University in St. Louis: "Learning to Compose Visual Relations" by Nan Liu et al (Nov 2021)
Leading researcher Pieter Abbeel (University of California, Berkeley) published "Zero-Shot Text-Guided Object Generation with Dream Fields", which had 24 shares over the past 2 days. @PhilRead tweeted "Coming to an architect near you: “Computer: Design a modern home for a family of five based on an aesthetic sourced from all publicly available social media profiles of my client suitable for entertaining along the coast of North Carolina for a budget of 400/sf.”".Thispaperwasalsosharedthemostonsocialmediawith128tweets.@PhilRead(PhilRead)tweeted"Comingtoanarchitectnearyou:“Computer:DesignamodernhomeforafamilyoffivebasedonanaestheticsourcedfromallpubliclyavailablesocialmediaprofilesofmyclientsuitableforentertainingalongthecoastofNorthCarolinaforabudgetof400/sf.”".Thispaperwasalsosharedthemostonsocialmediawith128tweets.@PhilRead(PhilRead)tweeted"Comingtoanarchitectnearyou:“Computer:DesignamodernhomeforafamilyoffivebasedonanaestheticsourcedfromallpubliclyavailablesocialmediaprofilesofmyclientsuitableforentertainingalongthecoastofNorthCarolinaforabudgetof400/sf.”".
This week was active for "Computer Science - Computers and Society", with 40 new papers.
The paper discussed most in the news over the past week was "Unique on Facebook: Formulation and Evidence of (Nano)targeting Individual Users with non-PII Data" by José González-Cabañas et al (Oct 2021), which was referenced 11 times, including in the article Digital Bridge: Transatlantic competition — Washington’s privacy plan — Microsoft lobbying in Politico.eu. The paper author, Angel Cuevas, was quoted saying "It is surprising to find that Facebook is implicitly recognizing that nanotargeting is feasible and the only countermeasure is assuming advertisers are unable to infer users interests". The paper got social media traction with 408 shares. The investigators define a data - driven model to quantify the number of interests from a user that make them unique on Facebook. A Twitter user, @WolfieChristl, commented "FB sorts billions into thousands of categories. This paper shows, based on 2017 experiments, that a single individual out of billions can be targeted by using only 4 of the rarest or 22 random 'interest' categories that FB assigns to users: HT".
This week was active for "Computer Science - Human-Computer Interaction", with 34 new papers.
This week was extremely active for "Computer Science - Learning", with 539 new papers.
The paper discussed most in the news over the past week was by a team at Massachusetts Institute of Technology: "ExoMiner: A Highly Accurate and Explainable Deep Learning Classifier to Mine Exoplanets" by Hamed Valizadegan et al (Nov 2021), which was referenced 81 times, including in the article AI discovers over 300 unknown exoplanets in Kepler telescope data in Tech Register. The paper author, Hamed Valizadegan (Machine learning manager with the Universities Space Research Association at Ames), was quoted saying "When ExoMiner says something is a planet, you can be sure it’s a planet. ExoMiner is highly accurate and in some ways more reliable than both existing machine classifiers and the human experts it’s meant to emulate because of the biases that come with human labeling. Now that we’ve trained ExoMiner using Kepler data, with a little fine-tuning, we can transfer that learning to other missions, including TESS, which we’re currently working on. There’s room to grow." The paper got social media traction with 40 shares. A Twitter user, @storybywill, posted "We didn't. The team who developed ExoMine did, and here is their paper", while @summarizedml commented "ExoMiner is an explainable and robust classifier for transit signal classification. 📄".
Leading researcher Pieter Abbeel (University of California, Berkeley) came out with "Count-Based Temperature Scheduling for Maximum Entropy Reinforcement Learning" @summarizedml tweeted "A state-based temperature scheduling approach for SoftQ-Learning, and instantiate a count-based soft Q-Learning algorithm for SQL 📄".
The paper shared the most on social media this week is by a team at Google: "Quantum advantage in learning from experiments" by Hsin-Yuan Huang et al (Dec 2021) with 238 shares. @quantumVerd (Guillaume Verdon) tweeted "In March 2020, we released Quantum and its whitepaper, betting big on the hypothesis that quantum advantage will likely be achieved with QML on quantum data. Amazing to this dream come to fruition <2 years later! Congrats to Robert, Michael & team!🎉".
The most influential Twitter user discussing papers is AK who shared "FQ-ViT: Fully Quantized Vision Transformer without Retraining" by Yang Lin et al (Nov 2021)
This week was active for "Computer Science - Multiagent Systems", with 26 new papers.
Over the past week, 22 new papers were published in "Computer Science - Neural and Evolutionary Computing".
The paper shared the most on social media this week is by a team at Université Paris-Saclay: "Long-range and hierarchical language predictions in brains and algorithms" by Charlotte Caucheteux et al (Nov 2021) with 119 shares. @summarizedml (SummarizedML) tweeted "We show that enhancing these models with long-range and hierarchical predictions improves their brain-mapping. 📄".
The most influential Twitter user discussing papers is AK who shared "FQ-ViT: Fully Quantized Vision Transformer without Retraining" by Yang Lin et al (Nov 2021)
This week was very active for "Computer Science - Robotics", with 72 new papers.
The paper discussed most in the news over the past week was by a team at Washington University in St. Louis: "Learning to Compose Visual Relations" by Nan Liu et al (Nov 2021)
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) came out with "SEAL: Self-supervised Embodied Active Learning using Exploration and 3D Consistency" The investigators explore how they can build upon the data and models of Internet images and use them to adapt to robot vision without requiring any extra labels. @summarizedml tweeted "We build and utilize 3D semantic maps to learn an active exploration policy and use them to adapt to robot vision without requiring any extra labels. 📄". This paper was also shared the most on social media with 75 tweets. @summarizedml (SummarizedML) tweeted "We build and utilize 3D semantic maps to learn an active exploration policy and use them to adapt to robot vision without requiring any extra labels. 📄".