Week Ending 7.5.2020
RESEARCH WATCH: 7.5.2020
This week was active for "Computer Science - Artificial Intelligence", with 99 new papers.
The paper discussed most in the news over the past week was "The NetHack Learning Environment" by Heinrich Küttler et al (Jun 2020), which was referenced 5 times, including in the article NetHack: Fast & Complex Learning Environment For Testing RL Agent Robustness & Generalization in SyncedReview.com. The paper got social media traction with 69 shares. The authors present the NetHack Learning Environment (NLE), a scalable, procedurally generated, stochastic, rich, and challenging environment for RL research based on the popular single - player terminal - based roguelike game, NetHack. On Twitter, @egrefen commented "We also provide nifty tools such as an agent analysis dashboard (with more coming). Contributions welcome! You can read more about NLE here: and download the env, tools, and agent code: Now go get that Amulet of Yendor! 🙂".
Leading researcher Abhinav Gupta (Carnegie Mellon University) published "Empirically Verifying Hypotheses Using Reinforcement Learning" The investigators formulate hypothesis verification as an RL problem. @yudapearl tweeted "RL vs. CBN. In Causality p. 24 a Causal Bayesian Network (CBN) is defined as a parsimonious representation of all possible interventions. This paper seems to reason in the space of interventions, not their CBN. Transparency is lost; what is gained?".
The paper shared the most on social media this week is by a team at Stanford University: "Similarity Search for Efficient Active Learning and Search of Rare Concepts" by Cody Coleman et al (Jun 2020) with 54 shares. @cgnorthcutt (Curtis G. Northcutt) tweeted "Strong work on scalable active learning by my good friend and colleague a SEAL of ML 😘".
This week was active for "Computer Science - Computer Vision and Pattern Recognition", with 274 new papers.
The paper discussed most in the news over the past week was "DeepFaceLab: A simple, flexible and extensible face swapping framework" by Ivan Perov et al (May 2020), which was referenced 6 times, including in the article Disney Is Getting Into the Deepfake Game in Yahoo! News. The paper got social media traction with 46 shares. The investigators detail the principles that drive the implementation of DeepFaceLab and introduce the pipeline of it, through which every aspect of the pipeline can be modified painlessly by users to achieve their customization purpose, and its noteworthy that DeepFaceLab could achieve results with high fidelity and indeed indiscernible by mainstream forgery detection approaches.
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) came out with "Object Goal Navigation using Goal-Oriented Semantic Exploration", which had 23 shares over the past 3 days. @towards_AI tweeted "Posted new paper on Object Goal Navigation describing our winning entry to the #CVPR2020 Habitat ObjectNav Challenge. Arxiv: Webpage: Our model also works in the real-world on a Locobot! 👇 w. A. Gupta".
The paper shared the most on social media this week is "Causal Discovery in Physical Systems from Videos" by Yunzhu Li et al (Jul 2020) with 159 shares. @DataScienceNIG (DataScienceNigeria) tweeted "AI can model how fabrics interact with videos. Welcome V-CDN from researchers at & used to extract structured keypoint-based representation from videos, understand relationships between components, & makes predictions. More".
The most influential Twitter user discussing papers is Danilo J. Rezende who shared "Conditional Set Generation with Transformers" by Adam R Kosiorek et al (Jun 2020) and said: "Cool work from and on set prediction It exploits the fact that gradients of invariant functions (with respect to orthogonal representations of a symmetry group) are equivariant maps to make a permutation equivariant set predictor".
This week was active for "Computer Science - Computers and Society", with 35 new papers.
The paper discussed most in the news over the past week was by a team at Stanford University: "Large image datasets: A pyrrhic win for computer vision?" by Vinay Uday Prabhu et al (Jun 2020), which was referenced 7 times, including in the article MIT Takes Down Popular AI Dataset Due to Racist, Misogynistic Content in Gizmodo. The paper got social media traction with 28 shares. The authors investigate problematic practices and consequences of large scale vision datasets. On Twitter, @clancynewyork commented "Thank you, and for exposing the horrors of this training dataset for ML, 80 Million Tiny Images, in use since 2008! Now, forced to be withdrawn! Paper here: [".
The paper shared the most on social media this week is "Distributed consent and its impact on privacy and observability in social networks" by Juniper Lovato et al (Jun 2020) with 109 shares. @big_data_kane (C. Brandon Ogbunu) tweeted "Back in the day my boy’s dad (who used to drop knowledge, give life lessons) said “Show me your friends, and I’ll tell you what you’re all about.” Well, he was right, and its a security issue in online space, say and colleagues in this dope work!".
Over the past week, 19 new papers were published in "Computer Science - Human-Computer Interaction".
The paper discussed most in the news over the past week was by a team at University of Colorado Boulder: "A Quantitative Portrait of Wikipedias High-Tempo Collaborations during the 2020 Coronavirus Pandemic" by Brian C. Keegan et al (Jun 2020), which was referenced 3 times, including in the article As the coronavirus spread, two social media communities drifted apart in Tech Xplore. The paper author, Brian Keegan, was quoted saying "You're seeing these online communities explore what works and what doesn't work when it comes to different ways of doing governance". The paper got social media traction with 28 shares. A user, @petersuber, tweeted "In the first 5 months of 2020, 134,337 contributors made 973,940 edits to 4,238 articles on #COVID19-related topics. Here's a study of that frenzied collaboration", while @WikiResearch observed ""A Quantitative Portrait of Wikipedia’s High-Tempo Collaborations during the 2020 Coronavirus Pandemic", based on the analysis of almost 1M revisions in English Wikipedia. (Keegan and Tan, 2020) #COVID #COVID19".
This week was extremely active for "Computer Science - Learning", with 531 new papers.
The paper discussed most in the news over the past week was "DeepFaceLab: A simple, flexible and extensible face swapping framework" by Ivan Perov et al (May 2020)
Leading researcher Yoshua Bengio (Université de Montréal) came out with "Object Files and Schemata: Factorizing Declarative and Procedural Knowledge in Dynamical Systems" @FrancescoLocat8 tweeted "I really enjoyed reading the new version of RIMs from et al. They consider object files that are temporally consistent which is great to disentangle objects. They have a very neat framework to model multiple instances of the same "class". Highly recommended read!".
The paper shared the most on social media this week is by a team at Google: "GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding" by Dmitry Lepikhin et al (Jun 2020) with 751 shares. @ankurbpn (Ankur Bapna) tweeted "600B parameter model, trains in 4 days on over 10 billion examples!! Amazing work on efficiently scaling up model capacities with sparsely-gated MoEs and SPMD by and others at".
The most influential Twitter user discussing papers is Danilo J. Rezende who shared "Conditional Set Generation with Transformers" by Adam R Kosiorek et al (Jun 2020)
Over the past week, ten new papers were published in "Computer Science - Multiagent Systems".
The paper discussed most in the news over the past week was "Simulating COVID-19 in a University Environment" by Philip T. Gressman et al (Jun 2020), which was referenced 89 times, including in the article Can College Campuses Prevent The Spread Of Covid-19 When They Reopen This Fall? in Forbes.com. The paper author, Jennifer Peck, was quoted saying "After all the academic and other precautions are taken, the reaction of students will be a big factor in determining what happens on campus". The paper also got the most social media traction with 150 shares. On Twitter, @DCBPhDV2 posted "Y'all. "In the absence of any intervention, all scenarios end with effectively all susceptible community members developing #COVID19 by the end of the semester, with peak infection rates reached between 20 and 40 days into the semester." #HigherEd".
Over the past week, 35 new papers were published in "Computer Science - Neural and Evolutionary Computing".
The paper discussed most in the news over the past week was "The NetHack Learning Environment" by Heinrich Küttler et al (Jun 2020)
Leading researcher Yoshua Bengio (Université de Montréal) came out with "Learning to Combine Top-Down and Bottom-Up Signals in Recurrent Neural Networks with Attention over Modules". This paper was also shared the most on social media with 37 tweets.
This week was very active for "Computer Science - Robotics", with 77 new papers.
The paper discussed most in the news over the past week was by a team at DeepMind: "dm_control: Software and Tasks for Continuous Control" by Yuval Tassa et al (Jun 2020), which was referenced 2 times, including in the article DeepMind Open-Sourced A Collection of Python Libraries For Augmenting RL Agents in Analytics India Magazine. The paper got social media traction with 6 shares. A user, @luxurytechappa1, tweeted "super paper that I have to re read".
Leading researcher Ruslan Salakhutdinov (Carnegie Mellon University) published "Object Goal Navigation using Goal-Oriented Semantic Exploration", which had 23 shares over the past 3 days. @towards_AI tweeted "Posted new paper on Object Goal Navigation describing our winning entry to the #CVPR2020 Habitat ObjectNav Challenge. Arxiv: Webpage: Our model also works in the real-world on a Locobot! 👇 w. A. Gupta". This paper was also shared the most on social media with 104 tweets. @towards_AI (Towards AI) tweeted "Posted new paper on Object Goal Navigation describing our winning entry to the #CVPR2020 Habitat ObjectNav Challenge. Arxiv: Webpage: Our model also works in the real-world on a Locobot! 👇 w. A. Gupta".