Abstract-Crowdsourcing has become an popular approach for annotating the large quantities of data required to train machine learning algorithms. However, obtaining labels in this manner poses two important challenges. First, naively labeling all of the data can be prohibitively expensive. Second, a significant fraction of the annotations can be incorrect due to carelessness or limited domain expertise of crowdsourced workers. Active learning provides a natural formulation to address the former issue by affordably selecting an appropriate subset of instances to label. Unfortunately, most active learning strategies are myopic and sensitive to label noise, which leads to poorly trained classifiers. We propose an active learning method that is specifically designed to be robust to such noise. We present an application of our technique in the domain of activity recognition for eldercare and validate the proposed approach using both simulated and realworld experiments using Amazon Mechanical Turk.

Today's service-oriented systems realize many ideas from the research conducted a decade or so ago in multiagent systems. Because these two fields are so deeply connected, further advances in multiagent systems could feed into tomorrow's successful service-oriented computing approaches.This article describes a 15-year roadmap for service-oriented multiagent system research. W e've already seen service-oriented computing (SOC) take hold in cross-enterprise business settings, such as the use of FedEx and UPS shipping services in e-commerce transactions; the aggregation of hotel, car rental, and airline services by Expedia and Orbitz; or bookrating services for libraries, consumers, and bookstores. Given the widespread interest in and deployment of Web services and service-oriented architectures that are occurring in industry, the scope of SOC in business settings will expand substantially. However, the emphasis has been on the execution of individual services and not on the more important problems of how services are selected and how they can collaborate to provide higher levels of functionality. Fortunately, four major trends in computing are addressing this problem:• Online ontologies are enabling meaning and understanding, arguably the last frontier for computing, to be captured and shared in more refined ways -via the Semantic Web initiative, for example, with the development of languages and representations for marking up heterogeneous content. In an alternative approach, shared representations are emerging from the works of (millions of) independent content developers. These ontologies will form models for numerous real-world entities and systems, as well as for the meanings of documents and content. • The widespread availability of many different types of sensors and effectors (including actuators and robotic devices) will enable online entities to not only become aware of the physical world, but also to manipulate, change, and control it.These trends are the new enablers that will drive SOC and multiagent system (MAS) research in the next decade and beyond. They portend an era in which complex systems will be modeled and simulated not just to understand them, but also to form predictions and interpretations that guide the monitoring and managing of them. SOC brings to the fore additional considerations, such as the necessity of modeling autonomous and heterogeneous components in uncertain and dynamic environments. Such components must be autonomously reactive and proactive yet able to interact flexibly with other components and environments. As a result, they're best thought of as agents, which collectively form MASs. Additionally, the key MAS concepts are reflected directly in those of SOC:• Multiagent SystemsThe history of MASs mirrors the history of computing in general. In the 1980s, distributed computing over LANs and advances in expert systems motivated the initial interest in distributed agents. Because the resulting systems functioned in single organizations, cooperation was the main focus. In the...

No abstract

Abstract-Most real-world social networks are inherently dynamic and composed of communities that are constantly changing in membership. As a result, recent years have witnessed increased attention toward the challenging problem of detecting evolving communities. This paper presents a gametheoretic approach for community detection in dynamic social networks in which each node is treated as a rational agent who periodically chooses from a set of predefined actions in order to maximize its utility function. The community structure of a snapshot emerges after the game reaches Nash equilibrium; the partitions and agent information are then transferred to the next snapshot. An evaluation of our method on two real world dynamic datasets (AS-Internet Routers Graph and ASOregon Graph) demonstrates that we are able to report more stable and accurate communities over time compared to the benchmark methods.

Communities are vehicles for efficiently disseminating news, rumors, and opinions in human social networks. Modeling information diffusion through a network can enable us to reach a superior functional understanding of the effect of network structures such as communities on information propagation. The intrinsic assumption is that form follows functionrational actors exercise social choice mechanisms to join communities that best serve their information needs. Particle Swarm Optimization (PSO) was originally designed to simulate aggregate social behavior; our proposed diffusion model, PSODM (Particle Swarm Optimization Diffusion Model) models information flow in a network by creating particle swarms for local network neighborhoods that optimize a continuous version of Holland's hyperplane-defined objective functions. In this paper, we show how our approach differs from prior modeling work in the area and demonstrate that it outperforms existing model-based community detection methods on several social network datasets.

With Partially Observable Markov Decision Processes (POMDPs), Intelligent Tutoring Systems (ITSs) can model individual learners from limited evidence and plan ahead despite uncertainty. However, POMDPs need appropriate representations to become tractable in ITSs that model many learner features, such as mastery of individual skills or the presence of specific misconceptions. This article describes two POMDP representations-state queues and observation chains-that take advantage of ITS task properties and let POMDPs scale to represent over 100 independent learner features. A real-world military training problem is given as one example. A human study (n = 14) provides initial validation for the model construction. Finally, evaluating the experimental representations with simulated students helps predict their impact on ITS performance. The compressed representations can model a wide range of simulated problems with instructional efficacy equal to lossless representations. With improved tractability, POMDP ITSs can accommodate more numerous or more detailed learner states and inputs.

Abstract-One important aspect of creating game bots is adversarial motion planning: identifying how to move to counter possible actions made by the adversary. In this paper, we examine the problem of opponent interception, in which the goal of the bot is to reliably apprehend the opponent. We present an algorithm for motion planning that couples planning and prediction to intercept an enemy on a partially-occluded Unreal Tournament map. Human players can exhibit considerable variability in their movement preferences and do not uniformly prefer the same routes. To model this variability, we use inverse reinforcement learning to learn a player-specific motion model from sets of example traces. Opponent motion prediction is performed using a particle filter to track candidate hypotheses of the opponent's location over multiple time horizons. Our results indicate that the learned motion model has a higher tracking accuracy and yields better interception outcomes than other motion models and prediction methods.

This paper presents a cost minimization approach to the problem of human behavior recognition. Using full-body motion capture data acquired from human subjects, our system recognizes the behaviors that a human subject is performing from a set of military maneuvers, based on the subject's motion type and proximity to landmarks. Low-level motion classification is performed using support vector machines (SVMs) and a hidden Markov Model (HMM); output from the classifier is used as an input feature for the behavior recognizer. Given the dynamic and highly reactive nature of the domain, our system must handle behavior sequences that are frequently interrupted and often interleaved. To recognize such behavior sequences, we employ dynamic programming in conjunction with a behavior transition cost function to efficiently select the most parsimonious explanation for the human's actions. We demonstrate that our system is robust to action classification errors and deviations by the human subject from the expected set of behaviors. Our approach is well suited for incorporation into synthetic agents that cooperate or compete against human subjects in virtual reality training environments.