** Abstracting complex 3D shapes with parsimonious part-based representations has been a long standing goal in computer vision. This paper presents a learning-based solution to this problem which goes beyond the traditional 3D cuboid representation by exploiting superquadrics as atomic elements. We demonstrate that superquadrics lead to more expressive 3D scene parses while being easier to learn than 3D cuboid representations. Moreover, we provide an analytical solution to the Chamfer loss which avoids the need for computational expensive reinforcement learning or iterative prediction. Our model learns to parse 3D objects into consistent superquadric representations without supervision. Results on various ShapeNet categories as well as the SURREAL human body dataset demonstrate the flexibility of our model in capturing fine details and complex poses that could not have been modelled using cuboids. **

** How much does having visual priors about the world (e.g. the fact that the world is 3D) assist in learning to perform downstream motor tasks (e.g. delivering a package)? We study this question by integrating a generic perceptual skill set (e.g. a distance estimator, an edge detector, etc.) within a reinforcement learning framework--see Figure 1. This skill set (hereafter mid-level perception) provides the policy with a more processed state of the world compared to raw images. We find that using a mid-level perception confers significant advantages over training end-to-end from scratch (i.e. not leveraging priors) in navigation-oriented tasks. Agents are able to generalize to situations where the from-scratch approach fails and training becomes significantly more sample efficient. However, we show that realizing these gains requires careful selection of the mid-level perceptual skills. Therefore, we refine our findings into an efficient max-coverage feature set that can be adopted in lieu of raw images. We perform our study in completely separate buildings for training and testing and compare against visually blind baseline policies and state-of-the-art feature learning methods. **

** An important facet of reinforcement learning (RL) has to do with how the agent goes about exploring the environment. Traditional exploration strategies typically focus on efficiency and ignore safety. However, for practical applications, ensuring safety of the agent during exploration is crucial since performing an unsafe action or reaching an unsafe state could result in irreversible damage to the agent. The main challenge of safe exploration is that characterizing the unsafe states and actions is difficult for large continuous state or action spaces and unknown environments. In this paper, we propose a novel approach to incorporate estimations of safety to guide exploration and policy search in deep reinforcement learning. By using a cost function to capture trajectory-based safety, our key idea is to formulate the state-action value function of this safety cost as a candidate Lyapunov function and extend control-theoretic results to approximate its derivative using online Gaussian Process (GP) estimation. We show how to use these statistical models to guide the agent in unknown environments to obtain high-performance control policies with provable stability certificates. **

** By the widespread popularity of electronic devices, the emergence of biometric technology has brought significant convenience to user authentication compared with the traditional password and mode unlocking. Among many biological characteristics, the face is a universal and irreplaceable feature that does not need too much cooperation and can significantly improve the user's experience at the same time. Face recognition is one of the main functions of electronic equipment propaganda. Hence it's virtually worth researching in computer vision. Previous work in this field has focused on two directions: converting loss function to improve recognition accuracy in traditional deep convolution neural networks (Resnet); combining the latest loss function with the lightweight system (MobileNet) to reduce network size at the minimal expense of accuracy. But none of these has changed the network structure. With the development of AutoML, neural architecture search (NAS) has shown excellent performance in the benchmark of image classification. In this paper, we integrate NAS technology into face recognition to customize a more suitable network. We quote the framework of neural architecture search which trains child and controller network alternately. At the same time, we mutate NAS by incorporating evaluation latency into rewards of reinforcement learning and utilize policy gradient algorithm to search the architecture automatically with the most classical cross-entropy loss. The network architectures we searched out have got state-of-the-art accuracy in the large-scale face dataset, which achieves 98.77% top-1 in MS-Celeb-1M and 99.89% in LFW with relatively small network size. To the best of our knowledge, this proposal is the first attempt to use NAS to solve the problem of Deep Face Recognition and achieve the best results in this domain. **

** Urban autonomous driving decision making is challenging due to complex road geometry and multi-agent interactions. Current decision making methods are mostly manually designing the driving policy, which might result in sub-optimal solutions and is expensive to develop, generalize and maintain at scale. On the other hand, with reinforcement learning (RL), a policy can be learned and improved automatically without any manual designs. However, current RL methods generally do not work well on complex urban scenarios. In this paper, we propose a framework to enable model-free deep reinforcement learning in challenging urban autonomous driving scenarios. We design a specific input representation and use visual encoding to capture the low-dimensional latent states. Several state-of-the-art model-free deep RL algorithms are implemented into our framework, with several tricks to improve their performance. We evaluate our method in a challenging roundabout task with dense surrounding vehicles in a high-definition driving simulator. The result shows that our method can solve the task well and is significantly better than the baseline. **

** In decision making problems for continuous state and action spaces, linear dynamical models are widely employed. Specifically, policies for stochastic linear systems subject to quadratic cost functions capture a large number of applications in reinforcement learning. Selected randomized policies have been studied in the literature recently that address the trade-off between identification and control. However, little is known about policies based on bootstrapping observed states and actions. In this work, we show that bootstrap-based policies achieve a square root scaling of regret with respect to time. We also obtain results on the accuracy of learning the model's dynamics. Corroborative numerical analysis that illustrates the technical results is also provided. **

** Deep neural networks have become commonplace in the domain of reinforcement learning, but are often expensive in terms of the number of parameters needed. While compressing deep neural networks has of late assumed great importance to overcome this drawback, little work has been done to address this problem in the context of reinforcement learning agents. This work aims at making first steps towards model compression in an RL agent. In particular, we compress networks to drastically reduce the number of parameters in them (to sizes less than 3% of their original size), further facilitated by applying a global max pool after the final convolution layer, and propose using Actor-Mimic in the context of compression. Finally, we show that this global max-pool allows for weakly supervised object localization, improving the ability to identify the agent's points of focus. **

Deep Reinforcement Learning for Online Offloading in Wireless Powered Mobile-Edge Computing Networks

** Wireless powered mobile-edge computing (MEC) has recently emerged as a promising paradigm to enhance the data processing capability of low-power networks, such as wireless sensor networks and internet of things (IoT). In this paper, we consider a wireless powered MEC network that adopts a binary offloading policy, so that each computation task of wireless devices (WDs) is either executed locally or fully offloaded to an MEC server. Our goal is to acquire an online algorithm that optimally adapts task offloading decisions and wireless resource allocations to the time-varying wireless channel conditions. This requires quickly solving hard combinatorial optimization problems within the channel coherence time, which is hardly achievable with conventional numerical optimization methods. To tackle this problem, we propose a Deep Reinforcement learning-based Online Offloading (DROO) framework that implements a deep neural network as a scalable solution that learns the binary offloading decisions from the experience. It eliminates the need of solving combinatorial optimization problems, and thus greatly reduces the computational complexity especially in large-size networks. To further reduce the complexity, we propose an adaptive procedure that automatically adjusts the parameters of the DROO algorithm on the fly. Numerical results show that the proposed algorithm can achieve near-optimal performance while significantly decreasing the computation time by more than an order of magnitude compared with existing optimization methods. For example, the CPU execution latency of DROO is less than $0.1$ second in a $30$-user network, making real-time and optimal offloading truly viable even in a fast fading environment. **

** Boltzmann exploration is widely used in reinforcement learning to provide a trade-off between exploration and exploitation. Recently, in (Cesa-Bianchi et al., 2017) it has been shown that pure Boltzmann exploration does not perform well from a regret perspective, even in the simplest setting of stochastic multi-armed bandit (MAB) problems. In this paper, we show that a simple modification to Boltzmann exploration, motivated by a variation of the standard doubling trick, achieves $O(K\log^{1+\alpha} T)$ regret for a stochastic MAB problem with $K$ arms, where $\alpha>0$ is a parameter of the algorithm. This improves on the result in (Cesa-Bianchi et al., 2017), where an algorithm inspired by the Gumbel-softmax trick achieves $O(K\log^2 T)$ regret. We also show that our algorithm achieves $O(\beta(G) \log^{1+\alpha} T)$ regret in stochastic MAB problems with graph-structured feedback, without knowledge of the graph structure, where $\beta(G)$ is the independence number of the feedback graph. Additionally, we present extensive experimental results on real datasets and applications for multi-armed bandits with both traditional bandit feedback and graph-structured feedback. In all cases, our algorithm performs as well or better than the state-of-the-art. **

** Consider a collaborative task that requires communication. Two agents are placed in an environment and must create a language from scratch in order to coordinate. Recent work has been interested in what kinds of languages emerge when deep reinforcement learning agents are put in such a situation, and in particular in the factors that cause language to be compositional-i.e. meaning is expressed by combining words which themselves have meaning. Evolutionary linguists have also studied the emergence of compositional language for decades, and they find that in addition to structural priors like those already studied in deep learning, the dynamics of transmitting language from generation to generation contribute significantly to the emergence of compositionality. In this paper, we introduce these cultural evolutionary dynamics into language emergence by periodically replacing agents in a population to create a knowledge gap, implicitly inducing cultural transmission of language. We show that this implicit cultural transmission encourages the resulting languages to exhibit better compositional generalization and suggest how elements of cultural dynamics can be further integrated into populations of deep agents. **

** Reinforcement learning (RL) agents optimize only the features specified in a reward function and are indifferent to anything left out inadvertently. This means that we must not only specify what to do, but also the much larger space of what not to do. It is easy to forget these preferences, since these preferences are already satisfied in our environment. This motivates our key insight: when a robot is deployed in an environment that humans act in, the state of the environment is already optimized for what humans want. We can therefore use this implicit preference information from the state to fill in the blanks. We develop an algorithm based on Maximum Causal Entropy IRL and use it to evaluate the idea in a suite of proof-of-concept environments designed to show its properties. We find that information from the initial state can be used to infer both side effects that should be avoided as well as preferences for how the environment should be organized. Our code can be found at https://github.com/HumanCompatibleAI/rlsp. **

** Within Reinforcement Learning, there is a growing collection of research which aims to express all of an agent's knowledge of the world through predictions about sensation, behaviour, and time. This work can be seen not only as a collection of architectural proposals, but also as the beginnings of a theory of machine knowledge in reinforcement learning. Recent work has expanded what can be expressed using predictions, and developed applications which use predictions to inform decision-making on a variety of synthetic and real-world problems. While promising, we here suggest that the notion of predictions as knowledge in reinforcement learning is as yet underdeveloped: some work explicitly refers to predictions as knowledge, what the requirements are for considering a prediction to be knowledge have yet to be well explored. This specification of the necessary and sufficient conditions of knowledge is important; even if claims about the nature of knowledge are left implicit in technical proposals, the underlying assumptions of such claims have consequences for the systems we design. These consequences manifest in both the way we choose to structure predictive knowledge architectures, and how we evaluate them. In this paper, we take a first step to formalizing predictive knowledge by discussing the relationship of predictive knowledge learning methods to existing theories of knowledge in epistemology. Specifically, we explore the relationships between Generalized Value Functions and epistemic notions of Justification and Truth. **

** Within Reinforcement Learning, there is a fledgling approach to conceptualizing the environment in terms of predictions. Central to this predictive approach is the assertion that it is possible to construct ontologies in terms of predictions about sensation, behaviour, and time---to categorize the world into entities which express all aspects of the world using only predictions. This construction of ontologies is integral to predictive approaches to machine knowledge where objects are described exclusively in terms of how they are perceived. In this paper, we ground the Pericean model of semiotics in terms of Reinforcement Learning Methods, describing Peirce's Three Categories in the notation of General Value Functions. Using the Peircean model of semiotics, we demonstrate that predictions alone are insufficient to construct an ontology; however, we identify predictions as being integral to the meaning-making process. Moreover, we discuss how predictive knowledge provides a particularly stable foundation for semiosis\textemdash the process of making meaning\textemdash and suggest a possible avenue of research to design algorithmic methods which construct semantics and meaning using predictions. **

** We present RL-VAE, a graph-to-graph variational autoencoder that uses reinforcement learning to decode molecular graphs from latent embeddings. Methods have been described previously for graph-to-graph autoencoding, but these approaches require sophisticated decoders that increase the complexity of training and evaluation (such as requiring parallel encoders and decoders or non-trivial graph matching). Here, we repurpose a simple graph generator to enable efficient decoding and generation of molecular graphs. **