Week Ending 8.16.2020
RESEARCH WATCH: 8.16.2020
Over the past week, 80 new papers were published in "Computer Science - Artificial Intelligence".
The paper discussed most in the news over the past week was "Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence" by Shakir Mohamed et al (Jul 2020), which was referenced 3 times, including in the article How to build an Ethical Data Science System without losing money? in Towards Data Science. The paper author, William Isaac, was quoted saying "It enables us a new grammar and vocabulary to talk about both why these issues matter and what we are going to do to think about and address these issues over the long run". The paper also got the most social media traction with 844 shares. The investigators explore the important role of critical science, and in particular of post - colonial and decolonial theories, in understanding and shaping the ongoing advances in artificial intelligence. A user, @natematias, tweeted "Encouraged to see more scholars connect postcolonial theory to tech. Nine years ago when I started my PhD, I quickly learned that CS wasn't interested in or even able to see that part of my expertise. Now that's changing".
Leading researcher Sergey Levine (University of California, Berkeley) came out with "Offline Meta-Reinforcement Learning with Advantage Weighting" The researchers study an analogous problem within reinforcement learning : can they enable an agent to leverage large, diverse experiences from previous tasks in order to quickly learn a new task?. @gastronomy tweeted "> Massive datasets have proven critical to successfully applying deep learning to real-world problems, catalyzing progress on tasks such as object recog".
The paper shared the most on social media this week is by a team at Microsoft: "Compression of Deep Learning Models for Text: A Survey" by Manish Gupta et al (Aug 2020) with 108 shares. @omarsar0 (elvis) tweeted "Lately, there have been efforts to "compress" deep learning models. Here is a recent survey paper on model compression particularly for deep learning-based NLP. Techniques covered: pruning, quantization, knowledge distillation, parameter sharing".
This week was active for "Computer Science - Computer Vision and Pattern Recognition", with 282 new papers.
The paper discussed most in the news over the past week was by a team at Google: "NeRF in the Wild: Neural Radiance Fields for Unconstrained Photo Collections" by Ricardo Martin-Brualla et al (Aug 2020), which was referenced 7 times, including in the article This AI Creates Detailed 3D Renderings from Thousands of Tourist Photos in PetaPixel. The paper got social media traction with 175 shares. On Twitter, @jbhuang0604 observed "Amazing results! Seems like NeRF has been around for ages, but it’s actually not been published at ECCV yet", while @EdwardDixon3 observed "In which a model learns geometry, texture and more from photos, disentangles scene lighting. The Trevi fountain is especially amazing, with the lighting and the water. !".
Leading researcher Pieter Abbeel (UC Berkeley) came out with "Visual Imitation Made Easy" The researchers present an alternate interface for imitation that simplifies the data collection process while allowing for easy transfer to robots. @dhadfieldmenell tweeted "This is some really cool work. Congrats".
The paper shared the most on social media this week is by a team at Seoul National University: "I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image" by Gyeongsik Moon et al (Aug 2020) with 177 shares.
This week was very active for "Computer Science - Computers and Society", with 45 new papers.
The paper discussed most in the news over the past week was "Decolonial AI: Decolonial Theory as Sociotechnical Foresight in Artificial Intelligence" by Shakir Mohamed et al (Jul 2020)
The paper shared the most on social media this week is "A Deep Learning Approach for COVID-19 Trend Prediction" by Tong Yang et al (Aug 2020) with 61 shares. The investigators developed a deep learning model - based approach to forecast the spreading trend of SARS - CoV-2 in the United States.
This week was very active for "Computer Science - Human-Computer Interaction", with 37 new papers.
The paper discussed most in the news over the past week was by a team at Microsoft: "Pen-based Interaction with Spreadsheets in Mobile Virtual Reality" by Travis Gesslein et al (Aug 2020), which was referenced 5 times, including in the article Spreadsheets are getting a 3D avatar in Hindustan Times. The paper author, Jens Grubert (Coburg University of Applied Sciences and Arts), was quoted saying "long time collaborators". The paper was shared 2 times in social media.
This week was very active for "Computer Science - Learning", with 308 new papers.
The paper discussed most in the news over the past week was "The Computational Limits of Deep Learning" by Neil C. Thompson et al (Jul 2020), which was referenced 17 times, including in the article Deep tech may stumble on insufficient computing power in Livemint.com. The paper author, Neil Thompson, was quoted saying "You have to throw a lot more computation at something to get a little improvement in performance". The paper got social media traction with 354 shares. A user, @DataScienceNIG, tweeted "Is Deep Learning reaching a computational limit? Maybe yes, check a work by team based on a review of 1,058 research papers, spanning 5 prominent application areas conducting computation per network pass & hardware burden analysis. More".
Leading researcher Yoshua Bengio (Université de Montréal) published "Mastering Rate based Curriculum Learning".
The paper shared the most on social media this week is by a team at Massachusetts Institute of Technology: "Sampling using $SU(N)$ gauge equivariant flows" by Denis Boyda et al (Aug 2020) with 395 shares. @seanjtaylor (Sean J. Taylor) tweeted "This title and abstract were super intimidating but Kyle did an amazing job of explaining the paper in the thread. Really neat to grok even a bit of cool projects like this one".
Over the past week, 18 new papers were published in "Computer Science - Multiagent Systems".
Over the past week, 16 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 New York University: "Self-supervised learning through the eyes of a child" by A. Emin Orhan et al (Jul 2020), which was referenced 3 times, including in the article Kid-mounted cameras help A.I. learn to view the world through eyes of a child in Digital Trends. The paper author, Emin Orhan, was quoted saying "The goal was to address a nature vs. nurture-type question". The paper got social media traction with 201 shares. On Twitter, @abecedarius posted "An exciting start on resolving those age-old questions about inductive bias in human learning (e.g. "poverty of the stimulus" in linguistics). They only used a week's worth of visual experience here, across the 2 years".
This week was active for "Computer Science - Robotics", with 61 new papers.
The paper discussed most in the news over the past week was by a team at Carnegie Mellon University: "Swoosh! Rattle! Thump! -- Actions that Sound" by Dhiraj Gandhi et al (Jul 2020), which was referenced 2 times, including in the article Exploring the interactions between sound, action and vision in robotics in Tech Xplore. The paper author, Lerrel Pinto (Carnegie Mellon University), was quoted saying "One exciting preliminary result of our study was that from sound alone you can recognize the type of object with close to 80% accuracy". The paper got social media traction with 8 shares. The researchers perform the first large - scale study of the interactions between sound and robotic action. A user, @SymbolicSound, tweeted "Study finds that sound can help a robot differentiate between object types & predict what action was applied to the object. Predictions derived from audio interactions were 24% better than those derived from passive visual representations".
Leading researcher Pieter Abbeel (UC Berkeley) published "Visual Imitation Made Easy" The researchers present an alternate interface for imitation that simplifies the data collection process while allowing for easy transfer to robots. @dhadfieldmenell tweeted "This is some really cool work. Congrats".
The paper shared the most on social media this week is "A Tendon-driven Robot Gripper with Passively Switchable Underactuated Surface and its Physics Simulation Based Parameter Optimization" by Tianyi Ko (Aug 2020) with 88 shares. The researchers propose a single - actuator gripper that can lift thin objects lying on a flat surface, in addition to the ability as a standard parallel gripper.