Week Ending 3.28.2021
RESEARCH WATCH: 3.28.2021
This week was active for "Computer Science", with 1,331 new papers.
The paper discussed most in the news over the past week was by a team at Carnegie Mellon University: "Gender Bias, Social Bias and Representation: 70 Years of Bollywood" by Kunal Khadilkar et al (Feb 2021), which was referenced 12 times, including in the article The Persistent Problem Of Cultural Discrimination In AI Datasets in Analytics India Magazine. The paper author, Ashiqur R. KhudaBukhsh (Carnegie Mellon University), was quoted saying "All of these things we kind of knew, but now we have numbers to quantify them". The paper got social media traction with 12 shares. On Twitter, @KunalKhadilkar posted "Hi Alison, I am so glad you found our research interesting. I would love to talk more about how we used linguistic techniques for uncovering social biases. The full paper is available at".
Leading researcher Pieter Abbeel (University of California, Berkeley) came out with "Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots".
The paper shared the most on social media this week is "Preliminary Analysis of Potential Harms in the Luca Tracing System" by Theresa Stadler et al (Mar 2021) with 737 shares. The investigators analyse the potential harms a large - scale deployment of the Luca system might cause to individuals, venues, and communities. @saschakiefer (Sascha Kiefer) tweeted "Interesting and concerning read. might be worth considering when you dig into comparing check-in apps as you mentioned in your last episode".
This week was very active for "Computer Science - Artificial Intelligence", with 199 new papers.
The paper discussed most in the news over the past week was by a team at Université de Montréal: "Towards Causal Representation Learning" by Bernhard Schölkopf et al (Feb 2021), which was referenced 8 times, including in the article Why AI struggles to grasp cause and effect in The Next Web. The paper also got the most social media traction with 451 shares. A user, @NalKalchbrenner, tweeted "Causality in ML is one of those slippery concepts that are hard to get a good grip on - a bit like the concepts of consciousness and perhaps truth. This paper makes an attempt 👇", while @YisongMiao commented "Haven't read, seems like very interesting! RT for self-arxiv. Thanks!".
Leading researcher Pieter Abbeel (University of California, Berkeley) published "Reinforcement Learning for Robust Parameterized Locomotion Control of Bipedal Robots".
The paper shared the most on social media this week is by a team at Saarland University: "MasakhaNER: Named Entity Recognition for African Languages" by David Ifeoluwa Adelani et al (Mar 2021) with 439 shares. @MasakhaneNLP (Masakhane) tweeted "We're SO excited to present the first large publicly available high quality dataset for NER in 10 African languages, bringing together a variety of stakeholders: language speakers, dataset curators, NLP practitioners, and evaluation experts💕🌍 💪 (1/n)".
This week was very active for "Computer Science - Computer Vision and Pattern Recognition", with 398 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "High-Performance Large-Scale Image Recognition Without Normalization" by Andrew Brock et al (Feb 2021), which was referenced 11 times, including in the article Google DeepMind’s NFNets Offers Deep Learning Efficiency in InfoQ. The paper got social media traction with 1102 shares. The investigators develop an adaptive gradient clipping technique which overcomes these instabilities, and design a significantly improved class of Normalizer - Free ResNets. A user, @sohamde_, tweeted "Releasing NFNets: SOTA on ImageNet. Without normalization layers! Code: This is the third paper in a series that began by studying the benefits of BatchNorm and ended by designing highly performant networks w/o it. A thread: 1/8".
Leading researcher Luc Van Gool (Computer Vision Laboratory) came out with "Unsupervised Robust Domain Adaptation without Source Data".
The paper shared the most on social media this week is "iMAP: Implicit Mapping and Positioning in Real-Time" by Edgar Sucar et al (Mar 2021) with 225 shares. @ducha_aiki (Dmytro Mishkin) tweeted "Wow, that was fast. NERF for SLAM. remember our discussion about NERF vs COLMAP?".
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 Carnegie Mellon University: "Gender Bias, Social Bias and Representation: 70 Years of BHollywood" by Kunal Khadilkar et al (Feb 2021), which was referenced 12 times, including in the article The Persistent Problem Of Cultural Discrimination In AI Datasets in Analytics India Magazine. The paper author, Ashiqur R. KhudaBukhsh (Carnegie Mellon University), was quoted saying "All of these things we kind of knew, but now we have numbers to quantify them". The paper got social media traction with 12 shares. On Twitter, @KunalKhadilkar posted "Hi Alison, I am so glad you found our research interesting. I would love to talk more about how we used linguistic techniques for uncovering social biases. The full paper is available at".
The paper shared the most on social media this week is "Preliminary Analysis of Potential Harms in the Luca Tracing System" by Theresa Stadler et al (Mar 2021)
Over the past week, 16 new papers were published in "Computer Science - Human-Computer Interaction".
This week was very active for "Computer Science - Learning", with 424 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "High-Performance Large-Scale Image Recognition Without Normalization" by Andrew Brock et al (Feb 2021)
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) published "Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification" @neuroecology tweeted "Learning directly from examples (RL without reward) "In contrast to standard RL, which uses reward functions, we aim to learn a policy that reaches states that are likely to solve the task"".
The paper shared the most on social media this week is "Generative Minimization Networks: Training GANs Without Competition" by Paulina Grnarova et al (Mar 2021) with 239 shares. @jm_alexia (Alexia Jolicoeur-Martineau) tweeted "Really cool paper to train adversarial games in a very stable non-adversarial way. Sadly, the computational cost is extremely high (2*k=20 times more gradient evaluations than a GAN with D_steps=1). Hopefully, this will be made more scalable in the future. This has potential".
This week was active for "Computer Science - Multiagent Systems", with 27 new papers.
The paper discussed most in the news over the past week was "The AI Arena: A Framework for Distributed Multi-Agent Reinforcement Learning" by Edward W. Staley et al (Mar 2021), which was referenced 1 time, including in the article Weekly review of Reinforcement Learning papers #1 in Medium.com. The paper got social media traction with 9 shares. A user, @gastronomy, tweeted "> Advances in reinforcement learning (RL) have resulted in recent breakthroughs in the application of artificial intelligence (AI) acros".
Over the past week, 20 new papers were published in "Computer Science - Neural and Evolutionary Computing".
The paper discussed most in the news over the past week was "Combinatorial optimization and reasoning with graph neural networks" by Quentin Cappart et al (Feb 2021), which was referenced 1 time, including in the article Distribution-Free Graph Kernels and Spatial Graph Neural Networks in Medium.com. The paper got social media traction with 126 shares. On Twitter, @y0b1byte commented "Section 3.3 is 🔥🔥🔥. And the whole paper is amazing. I was that excited when I read the original GN paper which sparkled my interest in GNNs", while @rcsaxe observed "Idk how many times I’m going to read this, or how long it’s going to take, but at a first glance it looks like an unparalleled resource on GNNs. Really excited to dig into it!".
This week was extremely active for "Computer Science - Robotics", with 137 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "Manipulator-Independent Representations for Visual Imitation" by Yuxiang Zhou et al (Mar 2021), which was referenced 2 times, including in the article DeepMind Proposes Manipulation-Independent Representations for Imitation of Behaviours Demonstrated by Previously Unseen Manipulator Morphologies in SyncedReview.com. The paper got social media traction with 6 shares. The investigators explore the possibility of third - person visual imitation of manipulation trajectories, only from vision and without access to actions, demonstrated by embodiments different to the ones of their imitating agent.
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) came out with "Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification" @neuroecology tweeted "Learning directly from examples (RL without reward) "In contrast to standard RL, which uses reward functions, we aim to learn a policy that reaches states that are likely to solve the task"".