We study the problem of object detection over scanned images of scientific documents. We consider images that contain objects of varying aspect ratios and sizes and range from coarse elements such as tables and figures to fine elements such as equations and section headers. We find that current object detectors fail to produce properly localized region proposals over such page objects. We revisit the original R-CNN model and present a method for generating fine-grained proposals over document elements. We also present a region embedding model that uses the convolutional maps of a proposal's neighbors as context to produce an embedding for each proposal. This region embedding is able to capture the semantic relationships between a target region and its surrounding context. Our end-to-end model produces an embedding for each proposal, then classifies each proposal by using a multi-head attention model that attends to the most important neighbors of a proposal. To evaluate our model, we collect and annotate a dataset of publications from heterogeneous journals. We show that our model, referred to as Attentive-RCNN, yields a 17% mAP improvement compared to standard object detection models.

Stochastic-sampling-based Generative Neural Networks, such as Restricted Boltzmann Machines and Generative Adversarial Networks, are now used for applications such as denoising, image occlusion removal, pattern completion, and motion synthesis. In scenarios which involve performing such inference tasks with these models, it is critical to determine metrics that allow for model selection and/or maintenance of requisite generative performance under pre-specified implementation constraints. In this paper, we propose a new metric for evaluating generative model performance based on $p$-values derived from the combined use of Maximum Mean Discrepancy (MMD) and permutation-based (PT-based) resampling, which we refer to as PT-MMD. We demonstrate the effectiveness of this metric for two cases: (1) Selection of bitwidth and activation function complexity to achieve minimum power-at-performance for Restricted Boltzmann Machines; (2) Quantitative comparison of images generated by two types of Generative Adversarial Networks (PGAN and WGAN) to facilitate model selection in order to maximize the fidelity of generated images. For these applications, our results are shown using Euclidean and Haar-based kernels for the PT-MMD two sample hypothesis test. This demonstrates the critical role of distance functions in comparing generated images against their corresponding ground truth counterparts as what would be perceived by human users.

In particular, we justify its use for fixed-confidence best-arm identification . We further propose a variant of TTTS called Top-Two Transportation Cost ( T3C), which disposes of the computational burden of TTTS . As our main contribution, we provide the first sample complexity analysis of TTTS and T3C when coupled with a very natural Bayesian stopping rule, for bandits with Gaussian rewards, solving one of the open questions raised by Russo (2016). We also provide new posterior convergence results for TTTS under two models that are commonly used in practice: bandits with Gaussian and Bernoulli rewards and conjugate priors. 1 Introduction In multi-armed bandits, a learner repeatedly chooses an arm to play, and receives a reward from the associated unknown probability distribution. When the task is best-arm identification (BAI), the learner is not only asked to sample an arm at each stage, but is also asked to output a recommendation (i.e., a guess for the arm with the largest mean reward) after a certain period.

Comparing data defined over space and time is notoriously hard, because it involves quantifying both spatial and temporal variability, while at the same time taking into account the chronological structure of data. Dynamic Time Warping (DTW) computes an optimal alignment between time series in agreement with the chronological order, but is inherently blind to spatial shifts. In this paper, we propose Spatio-Temporal Alignments (STA), a new differentiable formulation of DTW, in which spatial differences between time samples are accounted for using regularized optimal transport (OT). Our temporal alignments are handled through a smooth variant of DTW called soft-DTW, for which we prove a new property: soft-DTW increases quadratically with time shifts. The cost matrix within soft-DTW that we use are computed using unbalanced OT, to handle the case in which observations are not normalized probabilities. Experiments on handwritten letters and brain imaging data confirm our theoretical findings and illustrate the effectiveness of STA as a dissimilarity for spatio-temporal data.

We develop a progressive training approach for neural networks which adaptively grows the network structure by splitting existing neurons to multiple off-springs. By leveraging a functional steepest descent idea, we derive a simple criterion for deciding the best subset of neurons to split and a splitting gradient for optimally updating the off-springs. Theoretically, our splitting strategy is a second-order functional steepest descent for escaping saddle points in an $\infty$-Wasserstein metric space, on which the standard parametric gradient descent is a first-order steepest descent. Our method provides a new computationally efficient approach for optimizing neural network structures, especially for learning lightweight neural architectures in resource-constrained settings.

In the past decade, deep learning methods have seen tremendous success in various supervised and unsupervised learning tasks such as classification and generative modeling. More recently, deep neural networks have emerged in the domain of reinforcement learning as a tool to solve decision-making problems of unprecedented complexity, e.g., navigation problems or game-playing AI. Despite the successful combinations of ideas from quantum computing with machine learning methods, there have been relatively few attempts to design quantum algorithms that would enhance deep reinforcement learning. This is partly due to the fact that quantum enhancements of deep neural networks, in general, have not been as extensively investigated as other quantum machine learning methods. In contrast, projective simulation is a reinforcement learning model inspired by the stochastic evolution of physical systems that enables a quantum speed-up in decision making. In this paper, we develop a unifying framework that connects deep learning and projective simulation, opening the route to quantum improvements in deep reinforcement learning. Our approach is based on so-called generative energy-based models to design reinforcement learning methods with a computational advantage in solving complex and large-scale decision-making problems.

Human and AI are increasingly interacting and collaborating to accomplish various complex tasks in the context of diverse application domains (e.g., healthcare, transportation, and creative design). Two dynamic, learning entities (AI and human) have distinct mental model, expertise, and ability; such fundamental difference/mismatch offers opportunities for bringing new perspectives to achieve better results. However, this mismatch can cause unexpected failure and result in serious consequences. While recent research has paid much attention to enhancing interpretability or explainability to allow machine to explain how it makes a decision for supporting humans, this research argues that there is urging the need for both human and AI should develop specific, corresponding ability to interact and collaborate with each other to form a human-AI team to accomplish superior results. This research introduces a conceptual framework called "Co-Learning," in which people can learn with/from and grow with AI partners over time. We characterize three key concepts of co-learning: "mutual understanding," "mutual benefits," and "mutual growth" for facilitating human-AI collaboration on complex problem solving. We will present proof-of-concepts to investigate whether and how our approach can help human-AI team to understand and benefit each other, and ultimately improve productivity and creativity on creative problem domains. The insights will contribute to the design of Human-AI collaboration.

Significant progress has been made in the area of model-based reinforcement learning. State-of-the-art algorithms are now able to match the asymptotic performance of model-free methods while being significantly more data efficient. However, this success has come at a price: state-of-the-art model-based methods require significant computation interleaved with data collection, resulting in run times that take days, even if the amount of agent interaction might be just hours or even minutes. When considering the goal of learning in real-time on real robots, this means these state-of-the-art model-based algorithms still remain impractical. In this work, we propose an asynchronous framework for model-based reinforcement learning methods that brings down the run time of these algorithms to be just the data collection time. We evaluate our asynchronous framework on a range of standard MuJoCo benchmarks. We also evaluate our asynchronous framework on three real-world robotic manipulation tasks. We show how asynchronous learning not only speeds up learning w.r.t wall-clock time through parallelization, but also further reduces the sample complexity of model-based approaches by means of improving the exploration and by means of effectively avoiding the policy overfitting to the deficiencies of learned dynamics models.

Self-play methods based on regret minimization have become the state of the art for computing Nash equilibria in large two-players zero-sum extensive-form games. These methods fundamentally rely on the hierarchical structure of the players' sequential strategy spaces to construct a regret minimizer that recursively minimizes regret at each decision point in the game tree. In this paper, we introduce the first efficient regret minimization algorithm for computing extensive-form correlated equilibria in large two-player general-sum games with no chance moves. Designing such an algorithm is significantly more challenging than designing one for the Nash equilibrium counterpart, as the constraints that define the space of correlation plans lack the hierarchical structure and might even form cycles. We show that some of the constraints are redundant and can be excluded from consideration, and present an efficient algorithm that generates the space of extensive-form correlation plans incrementally from the remaining constraints. This structural decomposition is achieved via a special convexity-preserving operation that we coin scaled extension. We show that a regret minimizer can be designed for a scaled extension of any two convex sets, and that from the decomposition we then obtain a global regret minimizer. Our algorithm produces feasible iterates. Experiments show that it significantly outperforms prior approaches and for larger problems it is the only viable option.

-- Motivated by the stringent requirements of unstructured real-world where a plethora of unknown objects reside in arbitrary locations of the surface, we propose a voxel-based deep 3D Convolutional Neural Network (3D CNN) that generates feasible 6-DoF grasp poses in unrestricted workspace with reachability awareness. Unlike the majority of works that predict if a proposed grasp pose within the restricted workspace will be successful solely based on grasp pose stability, our approach further learns a reachability predictor that evaluates if the grasp pose is reachable or not from robot's own experience. T o avoid the laborious real training data collection, we exploit the power of simulation to train our networks on a large-scale synthetic dataset. This work is an early attempt that simultaneously evaluates grasping reachability from learned knowledge while proposing feasible grasp poses with 3D CNN. Experimental results in both simulation and real-world demonstrate that our approach outperforms several other methods and achieves 82.5% grasping success rate on unknown objects. I. INTRODUCTION Real-world applications demand robotic manipulation algorithms that are efficient in arbitrary workspace where objects may not be reachable. Figure 1 illustrates a scenario where such an algorithm needs to 1) decide which of the sampled grasp pose candidates are more reachable and 2) grasp as many objects as possible from the dense clutter with minimal efforts. The predominant top-down grasping is often restricted in narrowly prepared workspace [1], whereas practical problems are often in extended and obstacle-rich environments that require flexible 6-DoF grasp poses to reach objects. Albeit extensive researches have been conducted on this topic, the grasping reachability problem remains relatively unexplored.