The Kalman Filter (KF) is pervasively used to control a vast array of consumer, health and defense products. By grouping sets of symmetric state variables, the Relational Kalman Filter (RKF) enables us to scale the exact KF for large-scale dynamic systems. In this paper, we provide a parameter learning algorithm for RKF, and a regrouping algorithm that prevents the degeneration of the relational structure for efficient filtering. The proposed algorithms significantly expand the applicability of the RKFs by solving the following questions: (1) how to learn parameters for RKF from partial observations; and (2) how to regroup the degenerated state variables by noisy real-world observations. To our knowledge, this is the first paper on learning parameters in relational continuous probabilistic models. We show that our new algorithms significantly improve the accuracy and the efficiency of filtering large-scale dynamic systems.

Current computer involvement in adolescent social networks (youth between the ages of 11 and 17) provides new opportunities to study group dynamics, interactions amongst peers, and individual preferences. Nevertheless, most of the research in this area focuses on efficiently retrieving information that is explicit in large social networks (e.g., properties of the graph structure), but not on how to use the dynamics of the virtual social network to discover latent characteristics of the real-world social network. In this paper, we present the analysis of a game designed to take advantage of the familiarity of adolescents with online social networks, and describe how the data generated by the game can be used to identify bullies in 5th grade classrooms. We present a probabilistic model of the game and using the in-game interactions of the players (i.e., content of chat messages) infer their social role within their classroom (either a bully or non-bully). The evaluation of our model is done by using previously collected data from psychological surveys on the same 5th grade population and by comparing the performance of the new model with off-the-shelf classifiers.

We address the problem of making single-point decisions in large partially observable games, where players interleave observation, deliberation, and action. We present information set generation as a key operation needed to reason about games in this way. We show how this operation can be used to implement an existing decision-making algorithm. We develop a constraint satisfaction algorithm for performing information set generation and show that it scales better than the existing depth-first search approach on multiple non-trivial games.

Robotic manipulation is important for real, physical world applications. General Purpose manipulation with a robot (eg. delivering dishes, opening doors with a key, etc.) is demanding. It is hard because (1) objects are constrained in position and orientation, (2) many non-spatial constraints interact (or interfere) with each other, and (3) robots may have multi-degree of freedoms (DOF). In this paper we solve the problem of general purpose robotic manipulation using a novel combination of planning and motion planning. Our approach integrates motions of a robot with other (non-physical or external-to-robot) actions to achieve a goal while manipulating objects. It differs from previous, hierarchical approaches in that (a) it considers kinematic constraints in configuration space (C-space) together with constraints over object manipulations; (b) it automatically generates high-level (logical) actions from a C-space based motion planning algorithm; and (c) it decomposes a planning problem into small segments, thus reducing the complexity of planning.

Neural networks are relatively successful in recognizing individual patterns. However, when images consist of combination of patterns, a preprocessing step of segmentation is required to avoid combinatorial explosion of the training phase. In practical applications, segmentation is a context dependent task which itself requires recognition. In this paper we propose and develop a biologically inspired neural architecture that can recognize and count an arbitrary collection of objects even if trained with individual objects, without making use of additional segmentation algorithms. The two essential features that govern the neurons in this algorithm are 1. dynamical feedback and 2. competition for activation. We show analytically that while the equations governing the output neurons are highly nonlinear in individual feature amplitudes, they are linear in groups of feature amplitudes. We further demonstrate through simulations, that our architecture can precisely count and recognize scenes in which three and four non-overlapping patterns are presented simultaneously. The ability to generalize numerosity outside the training distribution with a simple learning scheme, lack of connection weights and segmentation algorithms prove regulatory feedback networks not only beneficial for machine learning tasks but also for biological modeling of animal vision.

Traditional connectionist models place an emphasis on learned weights. Based on neurobiological evidence, a new approach is developed and experimentally shown to be more robust for disambiguating novel combinations of stimuli. It does not require variable weights and avoids many training related issues. This approach is compared with traditional weight-learning methods. The network is better able to function in different scenarios and can recognize multiple stimuli even if it is only trained on single stimuli.

We present exact algorithms for identifying deterministic-actions' effects and preconditions in dynamic partially observable domains. They apply when one does not know the action model(the way actions affect the world) of a domain and must learn it from partial observations over time. Such scenarios are common in real world applications. They are challenging for AI tasks because traditional domain structures that underly tractability (e.g., conditional independence) fail there (e.g., world features become correlated). Our work departs from traditional assumptions about partial observations and action models. In particular, it focuses on problems in which actions are deterministic of simple logical structure and observation models have all features observed with some frequency. We yield tractable algorithms for the modified problem for such domains. Our algorithms take sequences of partial observations over time as input, and output deterministic action models that could have lead to those observations. The algorithms output all or one of those models (depending on our choice), and are exact in that no model is misclassified given the observations. Our algorithms take polynomial time in the number of time steps and state features for some traditional action classes examined in the AI-planning literature, e.g., STRIPS actions. In contrast, traditional approaches for HMMs and Reinforcement Learning are inexact and exponentially intractable for such domains. Our experiments verify the theoretical tractability guarantees, and show that we identify action models exactly. Several applications in planning, autonomous exploration, and adventure-game playing already use these results. They are also promising for probabilistic settings, partially observable reinforcement learning, and diagnosis.

Motion planning for robotic arms is important for real, physical world applications. The planning for arms with high-degree-of-freedom (DOF) is hard because its search space is large (exponential in the number of joints), and the links may collide with static obstacles or other joints (self-collision). In this paper we present a motion planning algorithm that finds plans of motion from one arm configuration to a goal arm configuration in 2D space assuming no self-collision. Our algorithm is unique in two ways: (a) it utilizes the topology of the arm and obstacles to factor the search space and reduce the complexity of the planning problem using dynamic programming; and (b) it takes only polynomial time in the number of joints under some conditions. We provide a sufficient condition for polytime motion planning for 2D-space arms: if there is a path between two homotopic configurations, an embedded local planner finds a path within a polynomial time. The experimental results show that the proposed algorithm improves the performance of path planning for 2D arms.

Over recent years, mobile Internet devices such as laptops, PDAs, smart phones etc, have become extremely popular and widespread. Once on board of a vehicle, these devices can automatically connect to the vehicle processor and thus greatly amplify the communications and processing capabilities available to the owner in a "pedestrian mode". We envision that this "amplification" opportunity will be one of the drivers of car to car and car to curb communications. In fact, the car communications system will not be used exclusively for mobile Internet access, but also as a distributed platform for the "opportunistic" cooperation among people with shared interests/goals. Exchanging safety messages among vehicles is a compelling example. Stretching opportunistic cooperation well beyond safety messages, we discuss in this paper the concept of virtual "flea market" over VANET called FleaNet In FleaNet, customers, either mobile (i.e., vehicles) or stationary (i.e., pedestrians, roadside shop owner), express their demands/offers, e.g., want to buy or sell an item, via radio queries. These queries are opportunistically disseminated exploiting in part the mobility of other customers in order to find the customer/vendor with matching needs/resources. In the paper we identify the key performance metrics, namely query resolution latency, scalability, and mobility. Based on the metrics, using models and simulation, we show that FleaNet can efficiently support a market place over vehicular networks

We devise the first constant factor approximation algorithm for minimum quotient vertex-cuts in planar graphs. Our algorithm achieves approximation ratio 1+4/3(1+ε) with running time O(W• n3+2/ε), where W is the total weight of the vertices. The approximation ratio improves to 4/3(1+ε+o(1)) if there is an optimal quotient vertex-cut (A*,B*,C*) where the weight of C* is of low order compared to those of A* and B*; this holds, for example, when the input graph has uniform weights and costs. The ratio further improves to 1+ε+o(1) if, in addition, min[w(A*),w(B*)] ≤ 1/3 W.We use our algorithm for quotient vertex-cuts to achieve the first constant-factor pseudo-approximation for vertex separators in planar graphs.Our technical contribution is two-fold. First, we prove a structural theorem for planar graphs, showing the existence of a near-optimal quotient vertex-cut whose high-level structure is that of a bounded-depth tree. Second, we develop an algorithm that optimizes over such complex structures in running time that depends (exponentially) not on the size of the structure, but rather only on its depth. These techniques may be applicable in other problems.