Week Ending 08.11.19
RESEARCH WATCH: 08.11.19
Over the past week, 600 new papers were published in "Computer Science".
The paper discussed most in the news over the past week was "Tracking sex: The implications of widespread sexual data leakage and tracking on porn websites" by Elena Maris et al (Jul 2019), which was referenced 198 times, including in the article If You Watch Porn, Facebook, Google, Oracle Are Tracking That in Moguldom. The paper author, Elena Maris (Postdoctoral researcher at Microsoft), was quoted saying "The fact that the mechanism for adult site tracking is so similar to, say, online retail should be a huge red flag. This isn’t picking out a sweater and seeing it follow you across the web. This is so much more specific and deeply personal". The paper got social media traction with 173 shares. The investigators explore tracking and privacy risks on pornography websites. On Twitter, @citadelo observed "This research focuses on porn-sites users' tracking: "analysis of 22,484 pornography websites indicated that 93% leak user data to a third party." Unfortunately even incognito window does not solve the entire problem. Another reason to use".
Leading researcher Pieter Abbeel (University of California, Berkeley) published "DoorGym: A Scalable Door Opening Environment And Baseline Agent" @hardmaru tweeted "DoorGym: A Scalable Door Opening Environment And Baseline Agent They introduce a highly configurable door simulation environment, as a step to move RL from toy environments towards atomic skills that can be composed and extended towards a broader goal".
The paper shared the most on social media this week is by a team at Georgia Institute of Technology: "ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks" by Jiasen Lu et al (Aug 2019) with 168 shares. @vykthur (Victor Dibia) tweeted "They show that learning joint representations of image content and natural language using a BERT based architecture (multi-modal two-stream model, co-attentional transformer layers) improves performance for image+language tasks such as visual question answering, etc".
Over the past week, 36 new papers were published in "Computer Science - Artificial Intelligence".
The paper discussed most in the news over the past week was "SentiMATE: Learning to play Chess through Natural Language Processing" by Isaac Kamlish et al (Jul 2019), which was referenced 5 times, including in the article Researchers explore natural language processing to assess chess moves in PhysOrg.com. The paper author, Isaac Kamlish, was quoted saying "SentiMATE: Learning to play Chess through Natural Language Processing". The paper got social media traction with 81 shares. A Twitter user, @jthteo, said "Wow. DeepLearning of chess by reading (NLP) rather than playing", while @Seanku observed "SentiMATE: Learning to play Chess ♟️ through Natural Language Processing. Developed ,It evaluates the quality of chess moves by analyzing the reaction of experts #DeepLearning #LSTM Paper".
Leading researcher Pieter Abbeel (University of California, Berkeley) published "DoorGym: A Scalable Door Opening Environment And Baseline Agent" @hardmaru tweeted
Over the past week, 158 new papers were published in "Computer Science - Computer Vision and Pattern Recognition".
The paper discussed most in the news over the past week was by a team at UC Berkeley: "Natural Adversarial Examples" by Dan Hendrycks et al (Jul 2019), which was referenced 19 times, including in the article Semantic Based Adversarial Examples Fool Face Recognition in SyncedReview.com. The paper author, Steven Basart, was quoted saying "Anyone willing to test their models against our data set is free to do so". The paper got social media traction with 538 shares. A user, @DanHendrycks, tweeted "Natural Adversarial Examples are real-world and unmodified examples which cause classifiers to be consistently confused. The new dataset has 7,500 images, which we personally labeled over several months. Paper: Dataset and code".
Leading researcher Dhruv Batra (Georgia Institute of Technology) came out with "ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks", which had 22 shares over the past 3 days. @vykthur tweeted "They show that learning joint representations of image content and natural language using a BERT based architecture (multi-modal two-stream model, co-attentional transformer layers) improves performance for image+language tasks such as visual question answering, etc". This paper was also shared the most on social media with 168 tweets. @vykthur (Victor Dibia) tweeted "They show that learning joint representations of image content and natural language using a BERT based architecture (multi-modal two-stream model, co-attentional transformer layers) improves performance for image+language tasks such as visual question answering, etc".
Over the past week, 20 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was "Tracking sex: The implications of widespread sexual data leakage and tracking on porn websites" by Elena Maris et al (Jul 2019)
This week was active for "Computer Science - Human-Computer Interaction", with 26 new papers.
The paper discussed most in the news over the past week was "I-Keyboard: Fully Imaginary Keyboard on Touch Devices Empowered by Deep Neural Decoder" by Ue-Hwan Kim et al (Jul 2019), which was referenced 5 times, including in the article Researchers Have Come up with an Invisible Keyboard for Touchscreens and VR in Beebom. The paper got social media traction with 15 shares. A Twitter user, @gastronomy, commented "> Text-entry aims to provide an effective and efficient pathway for humans to deliver their messages to computers. With the ad".
This week was active for "Computer Science - Learning", with 243 new papers.
The paper discussed most in the news over the past week was by a team at UC Berkeley: "Natural Adversarial Examples" by Dan Hendrycks et al (Jul 2019),
Leading researcher Pieter Abbeel (University of California, Berkeley) came out with "Dimensionality Reduction Flows" The authors propose methods to reduce the latent space dimension of flow models. @serrjoa tweeted "I've been thinking on how to reduce dimensionality in normalizing flows for a while now without success (involving matrix pseudo inverses and the like). Now I have something new to think on".
The paper shared the most on social media this week is by a team at DeepMind: "Behaviour Suite for Reinforcement Learning" by Ian Osband et al (Aug 2019) with 484 shares. The investigators introduce the Behaviour Suite for Reinforcement Learning, or bsuite for short. @koraykv (koray kavukcuoglu) tweeted "Open sourcing bsuite. Automated evaluation and analysis of agents on RL benchmarks. We hope this will help reproducible and accessible research on core problems in RL. Looking forward to seeing results from the community!".
Over the past week, 11 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".
This week was active for "Computer Science - Robotics", with 48 new papers.
The paper discussed most in the news over the past week was "Robot Learning of Shifting Objects for Grasping in Cluttered Environments" by Lars Berscheid et al (Jul 2019), which was referenced 1 time, including in the article An algorithm to teach robots pre-grasping manipulation strategies in PhysOrg.com. The paper author, Lars Berscheid(Researchers), was quoted saying "In addition, grasping objects is often part of a high-level task, e.g. we want to place the object at a specific position". The paper was shared 1 time in social media.
Leading researcher Pieter Abbeel (University of California, Berkeley) published "DoorGym: A Scalable Door Opening Environment And Baseline Agent" @hardmaru tweeted
The paper shared the most on social media this week is by a team at University of Vermont: "Word2vec to behavior: morphology facilitates the grounding of language in machines" by David Matthews et al (Aug 2019) with 102 shares. The authors introduce a method that does so by training robots to act similarly to semantically - similar word2vec encoded commands. @aqaderb (abu) tweeted "interesting research experiment: given accurate keypoint representations of individuals in a movie (audio-visual pairing), each step in a recurrent neural net could use the key points and the image frame to generate dialogue".
The most influential Twitter user discussing papers is Zachary Lipton who shared "Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift" by Stephan Rabanser et al (Oct 2018) and said: "Funny story, we discovered accidentally that MNIST is not i.i.d.!".