Abstract: We consider the problem of multiple agents sensing and acting in environments with the goal of maximising their shared utility. In these environments, agents must learn communication protocols in order to share information that is needed to solve the tasks. By embracing deep neural networks, we are able to demonstrate end-to-end learning of protocols in complex environments inspired by communication riddles and multi-agent computer vision problems with partial observability. We propose two approaches for learning in these domains: Reinforced Inter-Agent Learning (RIAL) and Differentiable Inter-Agent Learning (DIAL). The former uses deep Q-learning, while the latter exploits the fact that, during learning, agents can backpropagate error derivatives through (noisy) communication channels. Hence, this approach uses centralised learning but decentralised execution. Our experiments introduce new environments for studying the learning of communication protocols and present a set of engineering innovations that are essential for success in these domains.

Abstract: Machine learning (ML) is the fastest growing field in computer science, and health informatics is among the greatest challenges. The goal of ML is to develop algorithms which can learn and improve over time and can be used for predictions. Most ML researchers concentrate on automatic machine learning (aML), where great advances have been made, for example, in speech recognition, recommender systems, or autonomous vehicles. Automatic approaches greatly benefit from big data with many training sets. However, in the health domain, sometimes we are confronted with a small number of data sets or rare events, where aML-approaches suffer of insufficient training samples. Here interactive machine learning (iML) may be of help, having its roots in reinforcement learning, preference learning, and active learning. The term iML is not yet well used, so we define it as “algorithms that can interact with agents and can optimize their learning behavior through these interactions, where the agents can also be human.” This “human-in-the-loop” can be beneficial in solving computationally hard problems, e.g., subspace clustering, protein folding, or k-anonymization of health data, where human expertise can help to reduce an exponential search space through heuristic selection of samples. Therefore, what would otherwise be an NP-hard problem, reduces greatly in complexity through the input and the assistance of a human agent involved in the learning phase.

Abstract: We study the problem of perceiving forest or mountain trails from a single monocular image acquired from the viewpoint of a robot traveling on the trail itself. Previous literature focused on trail segmentation, and used low-level features such as image saliency or appearance contrast; we propose a different approach based on a deep neural network used as a supervised image classifier. By operating on the whole image at once, our system outputs the main direction of the trail compared to the viewing direction. Qualitative and quantitative results computed on a large real-world dataset (which we provide for download) show that our approach outperforms alternatives, and yields an accuracy comparable to the accuracy of humans that are tested on the same image classification task. Preliminary results on using this information for quadrotor control in unseen trails are reported. To the best of our knowledge, this is the first letter that describes an approach to perceive forest trials, which is demonstrated on a quadrotor micro aerial vehicle.

Abstract: Learning to navigate in complex environments with dynamic elements is an important milestone in developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency and task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. In particular we consider jointly learning the goal-driven reinforcement learning problem with auxiliary depth prediction and loop closure classification tasks. This approach can learn to navigate from raw sensory input in complicated 3D mazes, approaching human-level performance even under conditions where the goal location changes frequently. We provide detailed analysis of the agent behaviour, its ability to localise, and its network activity dynamics, showing that the agent implicitly learns key navigation abilities.

Abstract: Lifelong Machine Learning (or Lifelong Learning) is an advanced machine learning paradigm that learns continuously, accumulates the knowledge learned in previous tasks, and uses it to help future learning. In the process, the learner becomes more and more knowledgeable and effective at learning. This learning ability is one of the hallmarks of human intelligence. However, the current dominant machine learning paradigm learns in isolation: given a training dataset, it runs a machine learning algorithm on the dataset to produce a model. It makes no attempt to retain the learned knowledge and use it in future learning. Although this isolated learning paradigm has been very successful, it requires a large number of training examples, and is only suitable for well-defined and narrow tasks. In comparison, we humans can learn effectively with a few examples because we have accumulated so much knowledge in the past which enables us to learn with little data or effort. Lifelong learning aims to achieve this capability. As statistical machine learning matures, it is time to make a major effort to break the isolated learning tradition and to study lifelong learning to bring machine learning to new heights. Applications such as intelligent assistants, chatbots, and physical robots that interact with humans and systems in real-life environments are also calling for such lifelong learning capabilities. Without the ability to accumulate the learned knowledge and use it to learn more knowledge incrementally, a system will probably never be truly intelligent. This book serves as an introductory text and survey to lifelong learning.

Abstract: Understanding the 3D world is a fundamental problem in computer vision. However, learning a good representation of 3D objects is still an open problem due to the high dimensionality of the data and many factors of variation involved. In this work, we investigate the task of single-view 3D object reconstruction from a learning agent's perspective. We formulate the learning process as an interaction between 3D and 2D representations and propose an encoder-decoder network with a novel projection loss defined by the projective transformation. More importantly, the projection loss enables the unsupervised learning using 2D observation without explicit 3D supervision. We demonstrate the ability of the model in generating 3D volume from a single 2D image with three sets of experiments: (1) learning from single-class objects; (2) learning from multi-class objects and (3) testing on novel object classes. Results show superior performance and better generalization ability for 3D object reconstruction when the projection loss is involved.

Abstract: Applying end-to-end learning to solve complex, interactive, pixel-driven control tasks on a robot is an unsolved problem. Deep Reinforcement Learning algorithms are too slow to achieve performance on a real robot, but their potential has been demonstrated in simulated environments. We propose using progressive networks to bridge the reality gap and transfer learned policies from simulation to the real world. The progressive net approach is a general framework that enables reuse of everything from low-level visual features to high-level policies for transfer to new tasks, enabling a compositional, yet simple, approach to building complex skills. We present an early demonstration of this approach with a number of experiments in the domain of robot manipulation that focus on bridging the reality gap. Unlike other proposed approaches, our real-world experiments demonstrate successful task learning from raw visual input on a fully actuated robot manipulator. Moreover, rather than relying on model-based trajectory optimisation, the task learning is accomplished using only deep reinforcement learning and sparse rewards.

Abstract: The emerging field of quantum machine learning has the potential to substantially aid in the problems and scope of artificial intelligence. This is only enhanced by recent successes in the field of classical machine learning. In this work we propose an approach for the systematic treatment of machine learning, from the perspective of quantum information. Our approach is general and covers all three main branches of machine learning: supervised, unsupervised, and reinforcement learning. While quantum improvements in supervised and unsupervised learning have been reported, reinforcement learning has received much less attention. Within our approach, we tackle the problem of quantum enhancements in reinforcement learning as well, and propose a systematic scheme for providing improvements. As an example, we show that quadratic improvements in learning efficiency, and exponential improvements in performance over limited time periods, can be obtained for a broad class of learning problems.

Abstract: TensorFlow is a machine learning system that operates at large scale and in heterogeneous environments. Its computational model is based on dataflow graphs with mutable state. Graph nodes may be mapped to different machines in a cluster, and within each machine to CPUs, GPUs, and other devices. TensorFlow supports a variety of applications, but it particularly targets training and inference with deep neural networks. It serves as a platform for research and for deploying machine learning systems across many areas, such as speech recognition, computer vision, robotics, information retrieval, and natural language processing. In this talk, we describe TensorFlow and outline some of its applications. We also discuss the question of what TensorFlow and deep learning may have to do with functional programming. Although TensorFlow is not purely functional, many of its uses are concerned with optimizing functions (during training), then with applying those functions (during inference). These functions are defined as compositions of simple primitives (as is common in functional programming), with internal data representations that are learned rather than manually designed. TensorFlow is joint work with many other people in the Google Brain team and elsewhere. More information is available at tensorflow.org.

Abstract: Recent advances in technology are delivering robots of reduced size and cost. A natural outgrowth of these advances are systems comprised of large numbers of robots that collaborate autonomously in diverse applications. Research on effective autonomous control of such systems, commonly called swarms, has increased dramatically in recent years and received attention from many domains, such as bioinspired robotics and control theory. These kinds of distributed systems present novel challenges for the effective integration of human supervisors, operators, and teammates that are only beginning to be addressed. This paper is the first survey of human–swarm interaction (HSI) and identifies the core concepts needed to design a human–swarm system. We first present the basics of swarm robotics. Then, we introduce HSI from the perspective of a human operator by discussing the cognitive complexity of solving tasks with swarm systems. Next, we introduce the interface between swarm and operator and identify challenges and solutions relating to human–swarm communication, state estimation and visualization, and human control of swarms. For the latter, we develop a taxonomy of control methods that enable operators to control swarms effectively. Finally, we synthesize the results to highlight remaining challenges, unanswered questions, and open problems for HSI, as well as how to address them in future works.

Abstract: Due to a general shift in manufacturing paradigm from mass production towards mass customization, reconfigurable automation technologies, such as robots, are required. However, current industrial robot solutions are notoriously difficult to program, leading to high changeover times when new products are introduced by manufacturers. In order to compete on global markets, the factories of tomorrow need complete production lines, including automation technologies that can effortlessly be reconfigured or repurposed, when the need arises. In this paper we present the concept of general, self-asserting robot skills for manufacturing. We show how a relatively small set of skills are derived from current factory worker instructions, and how these can be transferred to industrial mobile manipulators. General robot skills can not only be implemented on these robots, but also be intuitively concatenated to program the robots to perform a variety of tasks, through the use of simple task-level programming methods. We demonstrate various approaches to this, extensively tested with several people inexperienced in robotics. We validate our findings through several deployments of the complete robot system in running production facilities at an industrial partner. It follows from these experiments that the use of robot skills, and associated task-level programming framework, is a viable solution to introducing robots that can intuitively and on the fly be programmed to perform new tasks by factory workers. HighlightsWe propose a conceptual model of robot skills and show how this differs from macros.We show how this approach can enable non-experts to utilize advanced robotic systems.Concrete industrial applications of the approach are presented, on advanced robot systems.

Abstract: New high-frequency multimodal data collection technologies and machine learning analysis techniques could offer new insights into learning, especially when students have the opportunity to generate unique, personalized artifacts, such as computer programs, robots, and solutions engineering challenges. To date most of the work on learning analytics and educational data mining has been focused on online courses and cognitive tutors, both of which provide a high degree of structure to the tasks, and are restricted to interactions that occur in front of a computer screen. In this paper, we argue that multimodal learning analytics can offer new insights into students’ learning trajectories in more complex and open-ended learning environments. We present several examples of this work and its educational application.

Abstract: Robots are becoming safe and smart enough to work alongside people not only on manufacturing production lines, but also in spaces such as houses, museums, or hospitals. This can be significantly exploited in situations in which a human needs the help of another person to perform a task, because a robot may take the role of the helper. In this sense, a human and the robotic assistant may cooperatively carry out a variety of tasks, therefore requiring the robot to communicate with the person, understand his/her needs, and behave accordingly. To achieve this, we propose a framework for a user to teach a robot collaborative skills from demonstrations. We mainly focus on tasks involving physical contact with the user, in which not only position, but also force sensing and compliance become highly relevant. Specifically, we present an approach that combines probabilistic learning, dynamical systems, and stiffness estimation to encode the robot behavior along the task. Our method allows a robot to learn not only trajectory following skills, but also impedance behaviors. To show the functionality and flexibility of our approach, two different testbeds are used: a transportation task and a collaborative table assembly.

Abstract: Though substantial research has been dedicated towards using technology to improve education, no current methods are as effective as one-on-one tutoring. A critical, though relatively understudied, aspect of effective tutoring is modulating the student's affective state throughout the tutoring session in order to maximize long-term learning gains. We developed an integrated experimental paradigm in which children play a second-language learning game on a tablet, in collaboration with a fully autonomous social robotic learning companion. As part of the system, we measured children's valence and engagement via an automatic facial expression analysis system. These signals were combined into a reward signal that fed into the robot's affective reinforcement learning algorithm. Over several sessions, the robot played the game and personalized its motivational strategies (using verbal and non-verbal actions) to each student. We evaluated this system with 34 children in preschool classrooms for a duration of two months. We saw that (1) children learned new words from the repeated tutoring sessions, (2) the affective policy personalized to students over the duration of the study, and (3) students who interacted with a robot that personalized its affective feedback strategy showed a significant increase in valence, as compared to students who interacted with a non-personalizing robot. This integrated system of tablet-based educational content, affective sensing, affective policy learning, and an autonomous social robot holds great promise for a more comprehensive approach to personalized tutoring.

Abstract: Machine learning, one of the top emerging sciences, has an extremely broad range of applications. However, many books on the subject provide only a theoretical approach, making it difficult for a newcomer to grasp the subject material. This book provides a more practical approach by explaining the concepts of machine learning algorithms and describing the areas of application for each algorithm, using simple practical examples to demonstrate each algorithm and showing how different issues related to these algorithms are applied.

Abstract: Reinforcement learning (RL) can automate a wide variety of robotic skills, but learning each new skill requires considerable real-world data collection and manual representation engineering to design policy classes or features. Using deep reinforcement learning to train general purpose neural network policies alleviates some of the burden of manual representation engineering by using expressive policy classes, but exacerbates the challenge of data collection, since such methods tend to be less efficient than RL with low-dimensional, hand-designed representations. Transfer learning can mitigate this problem by enabling us to transfer information from one skill to another and even from one robot to another. We show that neural network policies can be decomposed into "task-specific" and "robot-specific" modules, where the task-specific modules are shared across robots, and the robot-specific modules are shared across all tasks on that robot. This allows for sharing task information, such as perception, between robots and sharing robot information, such as dynamics and kinematics, between tasks. We exploit this decomposition to train mix-and-match modules that can solve new robot-task combinations that were not seen during training. Using a novel neural network architecture, we demonstrate the effectiveness of our transfer method for enabling zero-shot generalization with a variety of robots and tasks in simulation for both visual and non-visual tasks.

Abstract: Advances in technology and computing play an increasingly important role in the evolution of modern surgical techniques and paradigms. This article reviews the current role of machine learning (ML) techniques in the context of surgery with a focus on surgical robotics (SR). Also, we provide a perspective on the future possibilities for enhancing the effectiveness of procedures by integrating ML in the operating room. The review is focused on ML techniques directly applied to surgery, surgical robotics, surgical training and assessment. The widespread use of ML methods in diagnosis and medical image computing is beyond the scope of the review. Searches were performed on PubMed and IEEE Explore using combinations of keywords: ML, surgery, robotics, surgical and medical robotics, skill learning, skill analysis and learning to perceive. Studies making use of ML methods in the context of surgery are increasingly being reported. In particular, there is an increasing interest in using ML for developing tools to understand and model surgical skill and competence or to extract surgical workflow. Many researchers begin to integrate this understanding into the control of recent surgical robots and devices. ML is an expanding field. It is popular as it allows efficient processing of vast amounts of data for interpreting and real-time decision making. Already widely used in imaging and diagnosis, it is believed that ML will also play an important role in surgery and interventional treatments. In particular, ML could become a game changer into the conception of cognitive surgical robots. Such robots endowed with cognitive skills would assist the surgical team also on a cognitive level, such as possibly lowering the mental load of the team. For example, ML could help extracting surgical skill, learned through demonstration by human experts, and could transfer this to robotic skills. Such intelligent surgical assistance would significantly surpass the state of the art in surgical robotics. Current devices possess no intelligence whatsoever and are merely advanced and expensive instruments.

Abstract: This chapter surveys the main approaches developed to date to endow robots with the ability to learn from human guidance. The field is best known as robot programming by demonstration, robot learning from/by demonstration, apprenticeship learning and imitation learning. We start with a brief historical overview of the field. We then summarize the various approaches taken to solve four main questions: when, what, who and when to imitate. We emphasize the importance of choosing well the interface and the channels used to convey the demonstrations, with an eye on interfaces providing force control and force feedback. We then review algorithmic approaches to model skills individually and as a compound and algorithms that combine learning from human guidance with reinforcement learning. We close with a look on the use of language to guide teaching and a list of open issues.

Abstract: Current uses of robots in classrooms are reviewed and used to characterise four scenarios: (s1) Robot as Classroom Teacher; (s2) Robot as Companion and Peer; (s3) Robot as Care-eliciting Companion; and (s4) Telepresence Robot Teacher. The main ethical concerns associated with robot teachers are identified as: privacy; attachment, deception, and loss of human contact; and control and accountability. These are discussed in terms of the four identified scenarios. It is argued that classroom robots are likely to impact children's' privacy, especially when they masquerade as their friends and companions, when sensors are used to measure children's responses, and when records are kept. Social robots designed to appear as if they understand and care for humans necessarily involve some deception (itself a complex notion), and could increase the risk of reduced human contact. Children could form attachments to robot companions (s2 and s3), or robot teachers (s1) and this could have a deleterious effect on their social development. There are also concerns about the ability, and use of robots to control or make decisions about children's behaviour in the classroom. It is concluded that there are good reasons not to welcome fully fledged robot teachers (s1), and that robot companions (s2 and 3) should be given a cautious welcome at best. The limited circumstances in which robots could be used in the classroom to improve the human condition by offering otherwise unavailable educational experiences are discussed.

Abstract: What is the right supervisory signal to train visual representations? Current approaches in computer vision use category labels from datasets such as ImageNet to train ConvNets. However, in case of biological agents, visual representation learning does not require millions of semantic labels. We argue that biological agents use physical interactions with the world to learn visual representations unlike current vision systems which just use passive observations (images and videos downloaded from web). For example, babies push objects, poke them, put them in their mouth and throw them to learn representations. Towards this goal, we build one of the first systems on a Baxter platform that pushes, pokes, grasps and observes objects in a tabletop environment. It uses four different types of physical interactions to collect more than 130K datapoints, with each datapoint providing supervision to a shared ConvNet architecture allowing us to learn visual representations. We show the quality of learned representations by observing neuron activations and performing nearest neighbor retrieval on this learned representation. Quantitatively, we evaluate our learned ConvNet on image classification tasks and show improvements compared to learning without external data. Finally, on the task of instance retrieval, our network outperforms the ImageNet network on recall@1 by 3 %.

Abstract: This paper introduces a reinforcement learning method for exploring a corridor environment with the depth information from an RGB-D sensor only. The robot controller achieves obstacle avoidance ability by pre-training of feature maps using the depth information. The system is based on the recent Deep Q-Network (DQN) framework where a convolution neural network structure was adopted in the Q-value estimation of the Q-learning method. We separate the DQN into a supervised deep learning structure and a Q-learning network. The experiments of a Turtlebot in the Gazebo simulation environment show the robustness to different kinds of corridor environments. All of the experiments use the same pre-training deep learning structure. Note that the robot is traveling in environments which are different from the pre-training environment. It is the first time that raw sensor information is used to build such an exploring strategy for robotics by reinforcement learning.

Abstract: Tasks that require many sequential decisions or complex solutions are hard to solve using conventional reinforcement learning algorithms. Based on the semi Markov decision process setting (SMDP) and the option framework, we propose a model which aims to alleviate these concerns. Instead of learning a single monolithic policy, the agent learns a set of simpler sub-policies as well as the initiation and termination probabilities for each of those sub-policies. While existing option learning algorithms frequently require manual specification of components such as the sub-policies, we present an algorithm which infers all relevant components of the option framework from data. Furthermore, the proposed approach is based on parametric option representations and works well in combination with current policy search methods, which are particularly well suited for continuous real-world tasks. We present results on SMDPs with discrete as well as continuous state-action spaces. The results show that the presented algorithm can combine simple sub-policies to solve complex tasks and can improve learning performance on simpler tasks.

Abstract: This is the first text to provide a unified and self-contained introduction to visual pattern recognition and machine learning. It is useful as a general introduction to artifical intelligence and knowledge engineering, and no previous knowledge of pattern recognition or machine learning is necessary. Basic for various pattern recognition and machine learning methods. Translated from Japanese, the book also features chapter exercises, keywords, and summaries.

Abstract: Mobile robots that navigate in populated environments require the capacity to move efficiently, safely and in human-friendly ways. In this paper, we address this task using a learning approach that enables a mobile robot to acquire navigation behaviors from demonstrations of socially normative human behavior. In the past, such approaches have been typically used to learn only simple behaviors under relatively controlled conditions using rigid representations or with methods that scale poorly to large domains. We thus develop a flexible graph-based representation able to capture relevant task structure and extend Bayesian inverse reinforcement learning to use sampled trajectories from this representation. In experiments with a real robot and a large-scale pedestrian simulator, we are able to show that the approach enables a robot to learn complex navigation behaviors of varying degrees of social normativeness using the same set of simple features.

Abstract: SARSA, as one kind of on-policy reinforcement learning methods, is integrated with deep learning to solve the video games control problems in this paper. We use deep convolutional neural network to estimate the state-action value, and SARSA learning to update it. Besides, experience replay is introduced to make the training process suitable to scalable machine learning problems. In this way, a new deep reinforcement learning method, called deep SARSA is proposed to solve complicated control problems such as imitating human to play video games. From the experiments results, we can conclude that the deep SARSA learning shows better performances in some aspects than deep Q learning.

Abstract: Machine learning is one of the most important and successful techniques in contemporary computer science. It involves the statistical inference of models (such as classifiers) from data. It is often conceived in a very impersonal way, with algorithms working autonomously on passively collected data. However, this viewpoint hides considerable human work of tuning the algorithms, gathering the data, and even deciding what should be modeled in the first place. Examining machine learning from a human-centered perspective includes explicitly recognising this human work, as well as reframing machine learning workflows based on situated human working practices, and exploring the co-adaptation of humans and systems. A human-centered understanding of machine learning in human context can lead not only to more usable machine learning tools, but to new ways of framing learning computationally. This workshop will bring together researchers to discuss these issues and suggest future research questions aimed at creating a human-centered approach to machine learning.