Large multilayer neural networks trained with backpropagation have recently achieved state-of-the-art results in a wide range of problems. However, using backprop for neural net learning still has some disadvantages, e.g., having to tune a large number of hyperparameters to the data, lack of calibrated probabilistic predictions, and a tendency to overfit the training data. In principle, the Bayesian approach to learning neural networks does not have these problems. However, existing Bayesian techniques lack scalability to large dataset and network sizes. In this work we present a novel scalable method for learning Bayesian neural networks, called probabilistic backpropagation (PBP). Similar to classical backpropagation, PBP works by computing a forward propagation of probabilities through the network and then doing a backward computation of gradients. A series of experiments on ten real-world datasets show that PBP is significantly faster than other techniques, while offering competitive predictive abilities. Our experiments also show that PBP provides accurate estimates of the posterior variance on the network weights.

Interactive dynamic influence diagrams(I-DIDs) are a well recognized decision model that explicitly considers how multiagent interaction affects individual decision making. To predict behavior of other agents, I-DIDs require models of the other agents to be known ahead of time and manually encoded. This becomes a barrier to I-DID applications in a human-agent interaction setting, such as development of intelligent non-player characters(NPCs) in real-time strategy(RTS) games, where models of other agents or human players are often inaccessible to domain experts. In this paper, we use automatic techniques for learning behavior of other agents from replay data in RTS games. We propose a learning algorithm with improvement over existing work by building a full profile of agent behavior. This is the first time that data-driven learning techniques are embedded into the I-DID decision making framework. We evaluate the performance of our approach on two test cases.

Feature noising is an effective mechanism on reducing the risk of overfitting. To avoid an explosive searching space, existing work typically assumes that all features share a single noise level, which is often cross-validated. In this paper, we present a Bayesian feature noising model that flexibly allows for dimension-specific or group-specific noise levels, and we derive a learning algorithm that adaptively updates these noise levels. Our adaptive rule is simple and interpretable, by drawing a direct connection to the fitness of each individual feature or feature group. Empirical results on various datasets demonstrate the effectiveness on avoiding extensive tuning and sometimes improving the performance due to its flexibility.

We research and develop autonomous mobile service robots as Collaborative Robots, i.e., CoBots. For the last three years, our four CoBots have autonomously navigated in our multi-floor office buildings for more than 1,000km, as the result of the integration of multiple perceptual, cognitive, and actuations representations and algorithms. In this paper, we identify a few core aspects of our CoBots underlying their robust functionality. The reliable mobility in the varying indoor environments comes from a novel episodic non-Markov localization. Service tasks requested by users are the input to a scheduler that can consider different types of constraints, including transfers among multiple robots. With symbiotic autonomy, the CoBots proactively seek external sources of help to fill-in for their inevitable occasional limitations. We present sampled results from a deployment and conclude with a brief review of other features of our service robots.

This thesis focuses on the problem of temporal planning under uncertainty with explicit safety guarantees, which are enforced by means of chance constraints. We aim at elevating the level in which operators interact with autonomous agents and specify their desired behavior, while retaining a keen sensitivity to risk. Instead of relying on unconditional sequences, our goal is to allow contingent plans to be dynamically scheduled and conditioned on observations of the world while remaining safe. Contingencies add flexibility by allowing goals to be achieved through different methods, while observations allow the agent to adapt to the environment. We demonstrate the usefulness of our chance-constrained temporal planning approaches in real-world applications, such as partially observable power supply restoration and collaborative human-robot manufacturing.

In decision-theoretic planning problems, such as (partially observable) Markov decision problems or coordination graphs, agents typically aim to optimize a scalar value function. However, in many real-world problems agents are faced with multiple possibly conflicting objectives. In such multi-objective problems, the value is a vector rather than a scalar, and we need methods that compute a coverage set, i.e., a set of solutions optimal for all possible trade-offs between the objectives. In this project propose new multi-objective planning methods that compute the so-called convex coverage set (CCS): the coverage set for when policies can be stochastic, or the preferences are linear. We show that the CCS has favorable mathematical properties, and is typically much easier to compute that the Pareto front, which is often axiomatically assumed as the solution set for multi-objective decision problems.

Our research is within the area of artificial intelligence and multi-agent systems. More specifically, we focus on evaluating trust relationships between the agents in multi-agent e-marketplaces and sensor networks and aim to address the following problems: 1) how to identify a trustworthy (good quality) agent; 2) how to cope with dishonest advisors i.e., agents who provide misleading opinions about others.

Decentralized partially observable Markov decision processes (Dec-POMDPs) provide a general model for decision-making under uncertainty in cooperative decentralized settings, but are difficult to solve optimally (NEXP-Complete). As a new way of solving these problems, we recently introduced a method for transforming a Dec-POMDP into a continuous-state deterministic MDP with a piecewise-linear and convex value function. This new Dec-POMDP formulation, which we call an occupancy MDP, allows powerful POMDP and continuous-state MDP methods to be used for the first time. However, scalability remains limited when the number of agents or problem variables becomes large. In this paper, we show that, under certain separability conditions of the optimal value function, the scalability of this approach can increase considerably. This separability is present when there is locality of interaction between agents, which can be exploited to improve performance. Unlike most previous methods, the novel continuous-state MDP algorithm retains optimality and convergence guarantees. Results show that the extension using separability can scale to a large number of agents and domain variables while maintaining optimality.

Graphical models, such as Bayesian Networks and Markov networks play an important role in artificial intelligence and machine learning. Inference is a central problem to be solved on these networks. This, and other problems on these graph models are often known to be hard to solve in general, but tractable on graphs with bounded Treewidth. Therefore, finding or approximating the Treewidth of a graph is a fundamental problem related to inference in graphical models. In this paper, we study the approximability of a number of graph problems: Treewidth and Pathwidth of graphs, Minimum Fill-In, and a variety of different graph layout problems such as Minimum Cut Linear Arrangement. We show that, assuming Small Set Expansion Conjecture, all of these problems are NP-hard to approx- imate to within any constant factor in polynomial time.

We present an approach for cyberbullying detection based on state-of-the-art text classification and a common sense knowledge base, which permits recognition over a broad spectrum of topics in everyday life. We analyze a more narrow range of particular subject matter associated with bullying and construct BullySpace, a common sense knowledge base that encodes particular knowledge about bullying situations. We then perform joint reasoning with common sense knowledge about a wide range of everyday life topics. We analyze messages using our novel AnalogySpace common sense reasoning technique. We also take into account social network analysis and other factors. We evaluate the model on real-world instances that have been reported by users on Form spring, a social networking website that is popular with teenagers. On the intervention side, we explore a set of reflective user interaction paradigms with the goal of promoting empathy among social network participants. We propose an air traffic control-like dashboard, which alerts moderators to large-scale outbreaks that appear to be escalating or spreading and helps them prioritize the current deluge of user complaints. For potential victims, we provide educational material that informs them about how to cope with the situation, and connects them with emotional support from others. A user evaluation shows that in context, targeted, and dynamic help during cyberbullying situations fosters end-user reflection that promotes better coping strategies.