Week Ending 4.25.2021
RESEARCH WATCH: 4.25.2021
This week was active for "Computer Science", with 1,209 new papers.
The paper discussed most in the news over the past week was "Variational inference with a quantum computer" by Marcello Benedetti et al (Mar 2021), which was referenced 64 times, including in the article Giant leaps from small things - UK quantum firm sees reason in Diginomica. The paper author, Matthias Rosenkranz, was quoted saying "cannot offer simple explanations for their answers and struggle when asked how confident they are on certain possible outcomes". The paper got social media traction with 18 shares. A Twitter user, @rosenkranz, commented "Our new paper "Variational inference with a #quantum computer" has been out for a few days 🎉. We develop the methods, then demonstrate them using a few graphical models (e.g. hidden Markov). #QuantumComputing #MachineLearning".
Leading researcher Pieter Abbeel (University of California, Berkeley) published "VideoGPT: Video Generation using VQ-VAE and Transformers".
The paper shared the most on social media this week is by a team at Google: "Carbon Emissions and Large Neural Network Training" by David Patterson et al (Apr 2021) with 606 shares. @timnitGebru (Timnit Gebru) tweeted "This is like a never ending nightmare. Its just so unreal. A paper written by ~90% men authors (8 out 9 authors are men), one of whom fired co-authors from 100% underrepresented groups, for writing about environmental racism, sexism and other issues that affect us".
This week was very active for "Computer Science - Artificial Intelligence", with 217 new papers.
The paper discussed most in the news over the past week was by a team at University of California, Berkeley: "Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots" by Zhongyu Li et al (Mar 2021), which was referenced 10 times, including in the article AI is getting more life-like by copying a trick from human children in Inverse.com. The paper author, Yunzhu Li, was quoted saying "Compared to the videos released from Boston Dynamics, the bipedal Cassie robot operates at a low speed with relatively conservative motions on a plane, where the gap between simulation and the real world is relatively small". The paper got social media traction with 54 shares. A user, @svlevine, tweeted "A few folks pointed out to me that there are some factual errors in the MIT TR article. Certainly we are *not* claiming that our paper is the first to show RL for bipedal locomotion! Our prior work section covers lots of prior papers on this".
Leading researcher Sergey Levine (University of California, Berkeley) published "Outcome-Driven Reinforcement Learning via Variational Inference" The researchers discuss a new perspective on reinforcement learning, recasting it as the problem of inferring actions that achieve desired outcomes, rather than a problem of maximizing rewards. @gastronomy tweeted "> While reinforcement learning algorithms provide automated acquisition of optimal policies, practical application of such methods requires a number".
The paper shared the most on social media this week is by a team at Google: "VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Text" by Hassan Akbari et al (Apr 2021) with 284 shares. @lorenlugosch (Loren Lugosch) tweeted "My friend, if it extracts patches, it’s using convolutions. (Still looks nice though!)".
This week was very active for "Computer Science - Computer Vision and Pattern Recognition", with 326 new papers.
The paper discussed most in the news over the past week was by a team at Univ. Grenoble Alpes: "Self-supervised Pretraining of Visual Features in the Wild" by Priya Goyal et al (Mar 2021), which was referenced 9 times, including in the article Best of arXiv.org for AI, Machine Learning, and Deep Learning – March 2021 in InsideBIGDATA. The paper got social media traction with 182 shares. On Twitter, @artsiom_s commented "Facebook published its ultimate SElf-supERvised (SEER) model. - They pretrained it on a 1B random, unlabeled and uncurated Instagram images 👀. - SEER outperformed SOTA self-supervised systems, reaching 84.2% top-1 accuracy on ImageNet. 🛠️".
Leading researcher Pieter Abbeel (University of California, Berkeley) published "VideoGPT: Video Generation using VQ-VAE and Transformers".
The paper shared the most on social media this week is by a team at Google: "VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Text" by Hassan Akbari et al (Apr 2021)
Over the past week, 18 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was by a team at Google: "Carbon Emissions and Large Neural Network Training" by David Patterson et al (Apr 2021), which was referenced 1 time, including in the article #64: Google’s tips for reducing the CO2 emissions of training AI models in Medium.com. The paper got social media traction with 606 shares. A user, @timnitGebru, tweeted "This is like a never ending nightmare. Its just so unreal. A paper written by ~90% men authors (8 out 9 authors are men), one of whom fired co-authors from 100% underrepresented groups, for writing about environmental racism, sexism and other issues that affect us".
This week was active for "Computer Science - Human-Computer Interaction", with 30 new papers.
The paper discussed most in the news over the past week was by a team at UC Berkeley: "Auto-Tuned Sim-to-Real Transfer" by Yuqing Du et al (Apr 2021), which was referenced 2 times, including in the article Pieter Abbeel Team Proposes Task-Agnostic RL Method to Auto-Tune Simulations to the Real World in SyncedReview.com. The paper got social media traction with 52 shares. A user, @Jack_T_Collins, tweeted "Plenty of work coming out recently using an online approaches for sim2real where the method requires rollouts onto the physical robot and then improves the simulator. Very cool work".
This week was very active for "Computer Science - Learning", with 375 new papers.
The paper discussed most in the news over the past week was "Variational inference with a quantum computer" by Marcello Benedetti et al (Mar 2021)
Leading researcher Pieter Abbeel (University of California, Berkeley) published "VideoGPT: Video Generation using VQ-VAE and Transformers".
The paper shared the most on social media this week is by a team at Google: "Carbon Emissions and Large Neural Network Training" by David Patterson et al (Apr 2021)
Over the past week, eight new papers were published in "Computer Science - Multiagent Systems".
The paper discussed most in the news over the past week was by a team at Indian Institute of Technology Gandhinagar: "A Quasi-centralized Collision-free Path Planning Approach for Multi-Robot Systems" by Rohith G et al (Mar 2021), which was referenced 3 times, including in the article A technique to plan paths for multiple robots in flexible formations in Tech Xplore. The paper author, Madhu Vadali (Indian Institute of Technology Gandhinagar), was quoted saying "This quasi-centralized approach, wherein individual robots selfishly deviate from a centrally planned path to navigate without any collisions, is similar to practices and strategies found in nature, where agents in a formation move in a centrally planned path, but each agent selfishly ensures that it does not collide with obstacles in its environment". The paper was shared 3 times in social media. The researchers present a novel quasi - centralized approach for collision - free path planning of multi - robot systems (MRS) in obstacle - ridden environments. On Twitter, @iitgn posted "#IITGNResearchCapsule 33:Through their submissions to IEEE Robotics and Automation Letter & Robotics and Autonomous Systems Journal IITGN researchers led by Prof Madhu Vadali, present 2 novel strategies for path planning of MRS".
Over the past week, 29 new papers were published in "Computer Science - Neural and Evolutionary Computing".
This week was very active for "Computer Science - Robotics", with 90 new papers.
The paper discussed most in the news over the past week was by a team at University of California, Berkeley: "Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots" by Zhongyu Li et al (Mar 2021)
Leading researcher Sergey Levine (University of California, Berkeley) came out with "Reset-Free Reinforcement Learning via Multi-Task Learning: Learning Dexterous Manipulation Behaviors without Human Intervention".
The paper shared the most on social media this week is "Multi-Modal Fusion Transformer for End-to-End Autonomous Driving" by Aditya Prakash et al (Apr 2021) with 60 shares.