All Articles

Hot Papers 2020-07-28

1. Towards Learning Convolutions from Scratch

Behnam Neyshabur

Convolution is one of the most essential components of architectures used in computer vision. As machine learning moves towards reducing the expert bias and learning it from data, a natural next step seems to be learning convolution-like structures from scratch. This, however, has proven elusive. For example, current state-of-the-art architecture search algorithms use convolution as one of the existing modules rather than learning it from data. In an attempt to understand the inductive bias that gives rise to convolutions, we investigate minimum description length as a guiding principle and show that in some settings, it can indeed be indicative of the performance of architectures. To find architectures with small description length, we propose β\beta-LASSO, a simple variant of LASSO algorithm that, when applied on fully-connected networks for image classification tasks, learns architectures with local connections and achieves state-of-the-art accuracies for training fully-connected nets on CIFAR-10 (85.19%), CIFAR-100 (59.56%) and SVHN (94.07%) bridging the gap between fully-connected and convolutional nets.

2. Storywrangler: A massive exploratorium for sociolinguistic, cultural, socioeconomic, and political timelines using Twitter

Thayer Alshaabi, Jane L. Adams, Michael V. Arnold, Joshua R. Minot, David R. Dewhurst, Andrew J. Reagan, Christopher M. Danforth, Peter Sheridan Dodds

In real-time, Twitter strongly imprints world events, popular culture, and the day-to-day; Twitter records an ever growing compendium of language use and change; and Twitter has been shown to enable certain kinds of prediction. Vitally, and absent from many standard corpora such as books and news archives, Twitter also encodes popularity and spreading through retweets. Here, we describe Storywrangler, an ongoing, day-scale curation of over 100 billion tweets containing around 1 trillion 1-grams from 2008 to 2020. For each day, we break tweets into 1-, 2-, and 3-grams across 150+ languages, record usage frequencies, and generate Zipf distributions. We make the data set available through an interactive time series viewer, and as downloadable time series and daily distributions. We showcase a few examples of the many possible avenues of study we aim to enable including how social amplification can be visualized through ‘contagiograms’.

3. How Epidemic Psychology Works on Social Media: Evolution of responses to the COVID-19 pandemic

Luca Maria Aiello, Daniele Quercia, Ke Zhou, Marios Constantinides, Sanja Šćepanović, Sagar Joglekar

  • retweets: 26, favorites: 73 (07/29/2020 15:03:34)
  • links: abs | pdf
  • cs.CY | cs.SI

Disruptions resulting from an epidemic might often appear to amount to chaos but, in reality, can be understood in a systematic way through the lens of “epidemic psychology”. According to the father of this research field, Philip Strong, not only is the epidemic biological; there is also the potential for three social epidemics: of fear, moralization, and action. This work is the first study to empirically test Strong’s model at scale. It does so by studying the use of language on 39M social media posts in US about the COVID-19 pandemic, which is the first pandemic to spread this quickly not only on a global scale but also online. We identified three distinct phases, which parallel Kuebler-Ross’s stages of grief. Each of them is characterized by different regimes of the three social epidemics: in the refusal phase, people refused to accept reality despite the increasing numbers of deaths in other countries; in the suspended reality phase (started after the announcement of the first death in the country), people’s fear translated into anger about the looming feeling that things were about to change; finally, in the acceptance phase (started after the authorities imposed physical-distancing measures), people found a “new normal” for their daily activities. Our real-time operationalization of Strong’s model makes it possible to embed epidemic psychology in any real-time model (e.g., epidemiological and mobility models).

4. 3D Human Shape and Pose from a Single Low-Resolution Image with Self-Supervised Learning

Xiangyu Xu, Hao Chen, Francesc Moreno-Noguer, Laszlo A. Jeni, Fernando De la Torre

3D human shape and pose estimation from monocular images has been an active area of research in computer vision, having a substantial impact on the development of new applications, from activity recognition to creating virtual avatars. Existing deep learning methods for 3D human shape and pose estimation rely on relatively high-resolution input images; however, high-resolution visual content is not always available in several practical scenarios such as video surveillance and sports broadcasting. Low-resolution images in real scenarios can vary in a wide range of sizes, and a model trained in one resolution does not typically degrade gracefully across resolutions. Two common approaches to solve the problem of low-resolution input are applying super-resolution techniques to the input images which may result in visual artifacts, or simply training one model for each resolution, which is impractical in many realistic applications. To address the above issues, this paper proposes a novel algorithm called RSC-Net, which consists of a Resolution-aware network, a Self-supervision loss, and a Contrastive learning scheme. The proposed network is able to learn the 3D body shape and pose across different resolutions with a single model. The self-supervision loss encourages scale-consistency of the output, and the contrastive learning scheme enforces scale-consistency of the deep features. We show that both these new training losses provide robustness when learning 3D shape and pose in a weakly-supervised manner. Extensive experiments demonstrate that the RSC-Net can achieve consistently better results than the state-of-the-art methods for challenging low-resolution images.

5. From climate change to pandemics: decision science can help scientists have impact

Christopher M. Baker, Patricia T. Campbell, Iadine Chades, Angela J. Dean, Susan M. Hester, Matthew H. Holden, James M. McCaw, Jodie McVernon, Robert Moss, Freya M. Shearer, Hugh P. Possingham

Scientific knowledge and advances are a cornerstone of modern society. They improve our understanding of the world we live in and help us navigate global challenges including emerging infectious diseases, climate change and the biodiversity crisis. For any scientist, whether they work primarily in fundamental knowledge generation or in the applied sciences, it is important to understand how science fits into a decision-making framework. Decision science is a field that aims to pinpoint evidence-based management strategies. It provides a framework for scientists to directly impact decisions or to understand how their work will fit into a decision process. Decision science is more than undertaking targeted and relevant scientific research or providing tools to assist policy makers; it is an approach to problem formulation, bringing together mathematical modelling, stakeholder values and logistical constraints to support decision making. In this paper we describe decision science, its use in different contexts, and highlight current gaps in methodology and application. The COVID-19 pandemic has thrust mathematical models into the public spotlight, but it is one of innumerable examples in which modelling informs decision making. Other examples include models of storm systems (eg. cyclones, hurricanes) and climate change. Although the decision timescale in these examples differs enormously (from hours to decades), the underlying decision science approach is common across all problems. Bridging communication gaps between different groups is one of the greatest challenges for scientists. However, by better understanding and engaging with the decision-making processes, scientists will have greater impact and make stronger contributions to important societal problems.

6. Trick the Body Trick the Mind: Avatar representation affects the perception of available action possibilities in Virtual Reality

Tugce Akkoc, Emre Ugur, Inci Ayhan

  • retweets: 5, favorites: 77 (07/29/2020 15:03:35)
  • links: abs | pdf
  • cs.HC

In immersive Virtual Reality (VR), your brain can trick you into believing that your virtual hands are your real hands. Manipulating the representation of the body, namely the avatar, is a potentially powerful tool for the design of innovative interactive systems in VR. In this study, we investigated interactive behavior in VR by using the methods of experimental psychology. Objects with handles are known to potentiate the afforded action. Participants tend to respond faster when the handle is on the same side as the responding hand in bi-manual speed response tasks. In the first experiment, we successfully replicated this affordance effect in a Virtual Reality (VR) setting. In the second experiment, we showed that the affordance effect was influenced by the avatar, which was manipulated by two different hand types: 1) hand models with full finger tracking that are able to grasp objects, and 2) capsule-shaped — fingerless — hand models that are not able to grasp objects. We found that less than 5 minutes of adaptation to an avatar, significantly altered the affordance perception. Counter intuitively, action planning was significantly shorter with the hand model that is not able to grasp. Possibly, fewer action possibilities provided an advantage in processing time. The presence of a handle speeded up the initiation of the hand movement but slowed down the action completion because of ongoing action planning. The results were examined from a multidisciplinary perspective and the design implications for VR applications were discussed.

7. Statistical Bootstrapping for Uncertainty Estimation in Off-Policy Evaluation

Ilya Kostrikov, Ofir Nachum

In reinforcement learning, it is typical to use the empirically observed transitions and rewards to estimate the value of a policy via either model-based or Q-fitting approaches. Although straightforward, these techniques in general yield biased estimates of the true value of the policy. In this work, we investigate the potential for statistical bootstrapping to be used as a way to take these biased estimates and produce calibrated confidence intervals for the true value of the policy. We identify conditions - specifically, sufficient data size and sufficient coverage - under which statistical bootstrapping in this setting is guaranteed to yield correct confidence intervals. In practical situations, these conditions often do not hold, and so we discuss and propose mechanisms that can be employed to mitigate their effects. We evaluate our proposed method and show that it can yield accurate confidence intervals in a variety of conditions, including challenging continuous control environments and small data regimes.

8. Few-shot Knowledge Transfer for Fine-grained Cartoon Face Generation

Nan Zhuang, Cheng Yang

  • retweets: 15, favorites: 64 (07/29/2020 15:03:35)
  • links: abs | pdf
  • cs.CV

In this paper, we are interested in generating fine-grained cartoon faces for various groups. We assume that one of these groups consists of sufficient training data while the others only contain few samples. Although the cartoon faces of these groups share similar style, the appearances in various groups could still have some specific characteristics, which makes them differ from each other. A major challenge of this task is how to transfer knowledge among groups and learn group-specific characteristics with only few samples. In order to solve this problem, we propose a two-stage training process. First, a basic translation model for the basic group (which consists of sufficient data) is trained. Then, given new samples of other groups, we extend the basic model by creating group-specific branches for each new group. Group-specific branches are updated directly to capture specific appearances for each group while the remaining group-shared parameters are updated indirectly to maintain the distribution of intermediate feature space. In this manner, our approach is capable to generate high-quality cartoon faces for various groups.

Published 29 Jul 2020

ML Lead at Beatrust. (https://beatrust.com)Tatsuya Shirakawa on Twitter