The human race is facing one of the most meaningful public health emergencies in the modern era caused by the COVID-19 pandemic. This pandemic introduced various challenges, from lock-downs with significant economic costs to fundamentally altering the way of life for many people around the world. The battle to understand and control the virus is still at its early stages yet meaningful insights have already been made. The uncertainty of why some patients are infected and experience severe symptoms, while others are infected but asymptomatic, and others are not infected at all, makes managing this pandemic very challenging. Furthermore, the development of treatments and vaccines relies on knowledge generated from an ever evolving and expanding information space. Given the availability of digital data in the modern era, artificial intelligence (AI) is a meaningful tool for addressing the various challenges introduced by this unexpected pandemic. Some of the challenges include: outbreak prediction, risk modeling including infection and symptom development, testing strategy optimization, drug development, treatment repurposing, vaccine development, and others.

We introduce a domain specific language for procedural content generation (PCG) called Grammatical Item Generation Language (GIGL). GIGL supports a compact representation of PCG with stochastic grammars where generated objects maintain grammatical structures. Advanced features in GIGL allow flexible customizations of the stochastic generation process. GIGL is designed and implemented to have direct interface with C++, in order to be capable of integration into production games. We showcase the expressiveness and flexibility of GIGL on several representative problem domains in grammatical PCG, and show that the GIGL-based implementations run as fast as comparable C++ implementation and with less code.

We study the following connectivity formation problem: Robots equipped with radio transmitters with a bounded communication range are scattered over a large area. They would like to relocate so as to form a connected network as soon as possible. Where should each robot move? We present an $O(\sqrt{n})$-factor approximation algorithm for this problem when $n$ robots are initially distributed uniformly at random in a bounded area. In addition to analytical proofs, we verify the performance of our algorithm through simulations.

We consider the problem of covering an environment with a robot when the robot has limited energy budget. The environment is represented as a polygon with a grid, whose resolution is proportional to the robot size, imposed on it. There is a single charging station in the environment. At each time step, the robot can move from one grid cell to an adjacent one.The energy consumption when moving in the environment is assumed to be uniform and proportional to the distance traveled. Our goal is to minimize both the total distance and the number of visits to the charging station. We present a coverage path planning algorithm which has O(ln D) approxima-tion factor for both objectives, where D is the distance of thefurthest cell in the environment measured on the grid.

This paper discusses the current analysis of a large set of child-robot interaction experiments, in which a 2-4 year old child plays a set of games with a small humanoid robot, with the goal of detecting early signs of autism in toddlers based on the interactions. Our first goal in this paper is to condense and display these child-robot interactions as multi-channel time series, starting with the distances between the child, robot, parent, and experimenter.  Our second goal is to use these data displays to compare and contrast different children, with the aim of clustering children with similar interaction patterns.  Using a ceiling-mounted camera, we record the interaction between a child and a robot which performs different games and dances.  After analyzing the video footage for the locations of all people and the robot in the room over the variable length of the interaction, we condense the interactions into simplified, quantifiable, scale-invariant data.  We show how the distances between the child and robot, experimenter, and caregiver can be discretized into a few location zones and compared across children using classic similarity measures.  Proxemics (social distances) between the child, robot, caregiver, and experimenter during a child-robot interaction show differences between participants and hence can provide additional information for behavior analysis.

Epanechnikov Mean Shift is a simple yet empirically very effective algorithm for clustering. It localizes the centroids of data clusters via estimating modes of the probability distribution that generates the data points, using the "optimal" Epanechnikov kernel density estimator. However, since the procedure involves non-smooth kernel density functions,the convergence behavior of Epanechnikov mean shift lacks theoretical support as of this writing---most of the existing analyses are based on smooth functions and thus cannot be applied to Epanechnikov Mean Shift. In this work, we first show that the original Epanechnikov Mean Shift may indeed terminate at a non-critical point, due to the non-smoothness nature. Based on our analysis, we propose a simple remedy to fix it. The modified Epanechnikov Mean Shift is guaranteed to terminate at a local maximum of the estimated density, which corresponds to a cluster centroid, within a inite number of iterations. We also propose a way to avoid running the Mean Shift iterates from every data point, while maintaining good clustering accuracies under non-overlapping spherical Gaussian mixture models. This further pushes Epanechnikov Mean Shift to handle very large and high-dimensional data sets. Experiments show surprisingly good performance compared to the Lloyd's K-means algorithm and the EM algorithm.

Imitation learning refers to the problem where an agent learns a policy that mimics the demonstration provided by the expert, without any information on the cost function of the environment. Classical approaches to imitation learning usually rely on a restrictive class of cost functions that best explains the expert's demonstration, exemplified by linear functions of pre-defined features on states and actions. We show that the kernelization of a classical algorithm naturally reduces the imitation learning to a distribution learning problem, where the imitation policy tries to match the state-action visitation distribution of the expert. Closely related to our approach is the recent work on leveraging generative adversarial networks (GANs) for imitation learning, but our reduction to distribution learning is much simpler, robust to scarce expert demonstration, and sample efficient. We demonstrate the effectiveness of our approach on a wide range of high-dimensional control tasks.

Topic modeling enables exploration and compact representation of a corpus. The CaringBridge (CB) dataset is a massive collection of journals written by patients and caregivers during a health crisis. Topic modeling on the CB dataset, however, is challenging due to the asynchronous nature of multiple authors writing about their health journeys. To overcome this challenge we introduce the Dynamic Author-Persona topic model (DAP), a probabilistic graphical model designed for temporal corpora with multiple authors. The novelty of the DAP model lies in its representation of authors by a persona---where personas capture the propensity to write about certain topics over time. Further, we present a regularized variational inference (RVI) algorithm, which we use to encourage the DAP model's personas to be distinct. Our results show significant improvements over competing topic models---particularly after regularization, and highlight the DAP model's unique ability to capture common journeys shared by different authors.

Animating digital characters has an important role in computer assisted experiences, from video games to movies to interactive robotics. A critical challenge in the field is to generate animations which accurately reflect the state of the animated characters, without looking repetitive or unnatural. In this work, we investigate the problem of procedurally generating a diverse variety of facial animations that express a given semantic quality (e.g., very happy). To that end, we introduce a new learning heuristic called Precision Variety Learning (PVL) which actively identifies and exploits the fundamental trade-off between precision (how accurate positive labels are) and variety (how diverse the set of positive labels is). We both identify conditions where important theoretical properties can be guaranteed, and show good empirical performance in variety of conditions. Lastly, we apply our PVL heuristic to our motivating problem of generating smile animations, and perform several user studies to validate the ability of our method to produce a perceptually diverse variety of smiles for different target intensities.

Reports of the Workshops of the Thirty-First AAAI Conference on Artificial Intelligence