Revising Markov boundary for multiagent probabilistic inference

Error handling in multimodal voice-enabled interfaces of tour-guide robots using graphical models

This paper first analyzes systematically two classical exact inference algorithms for local inference in multiply sectioned Bayesian networks (MSBN) and points out the factor determining the complexity of the algorithms. Furthermore, the paper proves the identity of the two algorithms, gives a unified explanation for them and finds the class of Bayesian networks in which exact inference can be performed. Finally, the paper discusses how to reduce the complexity of the global inference in MSBN and gives some basic principles to guarantee the efficiency of the whole inference.

The interaction between natural events and technological installations involves complex mechanisms which have the potential to affect simultaneously more critical systems, nullifying the redundancy measures common to industrial safety systems and endangering the integrity of facilities. The concerns related to this kind of events are far from being restricted to a merely economic or industrial nature. On the contrary, due to the sensitivity of most processes performed in industrial plants and the negative consequences of eventual releases of hazardous materials, the impact of simultaneous failures embraces also the environment and population surrounding the installations. The risk is further widened by the trend of climate extremes: both observations over the past century and projections for next decades suggest an increase of the severity of extreme weather events and their frequency, both on local and global scales. The rise of sea water levels together with the exacerbation of extreme winds and precipitations, enlarge the geographic area of risk and rise the likelihood of accidents in regions historically susceptible to natural hazards. The prevention of technological accidents triggered by natural hazards lies unavoidably with the development of efficient theoretical and computational tools for the vulnerability assessment of industrial installations and the identification of effective strategies to tackle the growing risks to which they are subject. In spite of the increasing trend of the risk and the high-impact consequences, the current scientific literature still lacks robust means to tackle these issues effectively. The research presented in this dissertation addresses the critical need for novel theoretical and computational methods tailored for the risk assessment of complex systems threaten by extreme natural events. The specific requirements associated with the modelling of the interaction between external hazards and engineering systems have been determined, resulting in the identification of two main bottlenecks. On the one hand, this kind of analysis has to deal with the difficulty of representing accurately the complexity of technological systems and the mutual influence among their subsystems. On the other, the high degree of uncertainty affecting climate variables (due to their inner aleatory nature and the restricted information generally available), strongly bounds the accuracy and credibility of the results on which risk-informed decisions must be made. In this work, well-known traditional approaches (such as Bayesian Networks, Monte Carlo methods etc.) as well as cutting-edge methods from different sectors of the scientific literature have been adopted and integrated in order to obtain a novel theoretical strategy and computational tool able to overcome the limitations of the current state of the art. The result of the research is a complete tool for risk assessment and decision making support, based on the use of probabilistic graphical models and able to fully represent a wide spectrum of variables types and their uncertainty, to provide the implementation of flexible computational models as well as their computation and uncertainty quantification.

Soft evidential update for probabilistic multiagent systems

Probabilistic programming refers to the idea of using standard programming constructs for specifying probabilistic models from machine learning and statistics, and employing generic inference algorithms for answering various queries on these models, such as posterior inference and estimation of model evidence. Although this idea itself is not new and was, in fact, explored by several programming-language and statistics researchers in the early 2000, it is only in the last few years that probabilistic programming has gained a large amount of attention among researchers in machine learning and programming languages, and that expressive and efficient probabilistic programming systems (such as Anglican, Church, Figaro, Infer.net, PSI, PyMC, Stan, and Venture) started to appear and have been taken up by a nontrivial number of users. The primary goal of my talk is to introduce probabilistic programming to the CONCUR/QUEST/FORMATS audience. At the end of my talk, I want the audience to understand basic results and techniques in probabilistic programming and to feel that these results and techniques are relevant or at least related to what she or he studies, although they typically come from foreign research areas, such as machine learning and statistics. My talk will contain both technical materials and lessons that I learnt from my machine-learning colleagues in Oxford, who are developing a highly-expressive higher-order probabilistic programming language, called Anglican. It will also include my work on the denotational semantics of higher-order probabilistic programming languages and their inference algorithms, which are jointly pursued with colleagues in Cambridge, Edinburgh, Oxford and Tubingen.

As genomics becomes increasingly relevant to medicine, medical informatics and bioinformatics are gradually converging into a larger field that we call computational biomedicine. Developing a computational framework that is common to the different disciplines that compose computational biomedicine will be a major enabler of the further development and integration of this research domain. Probabilistic graphical models such as Hidden Markov Models, belief networks, and missing-data models together with computational methods such as dynamic programming, Expectation-Maximization, data-augmentation Gibbs sampling, and the Metropolis-Hastings algorithm provide the tools for an integrated probabilistic approach to computational biomedicine. We show how graphical models have already found a broad application in different fields composing computational biomedicine. We also indicate several challenges that lie at the interface between medical informatics, statistical genomics, and bioinformatics. We also argue that graphical models offer a unified framework making it possible to integrate in a statistically meaningful way multiple models ranging from the molecular level to cellular and to clinical levels. Because of their versatility and firm statistical underpinning, we assert that probabilistic graphical models can serve as the lingua franca for many computationally intensive approaches to biology and medicine. As such, graphical models should be a foundation of the curriculum of students in these fields. From such a foundation, students could then build towards specific computational methods in medical informatics, medical image analysis, statistical genetics, or bioinformatics while keeping the communication open between these areas.

A probabilistic graphical model approach to stochastic multiscale partial differential equations

Probabilistic graphical models are a fundamental tool in probabilistic modeling, machine learning and artificial intelligence. They allow us to integrate in a natural way expert knowledge, physical modeling, heterogeneous and correlated data and quantities of interest. For exactly this reason, multiple sources of model uncertainty are inherent within the modular structure of the graphical model. In this paper we develop information-theoretic, robust uncertainty quantification methods and non-parametric stress tests for directed graphical models to assess the effect and the propagation through the graph of multi-sourced model uncertainties to quantities of interest. These methods allow us to rank the different sources of uncertainty and correct the graphical model by targeting its most impactful components with respect to the quantities of interest. Thus, from a machine learning perspective, we provide a mathematically rigorous approach to correctability that guarantees a systematic selection for improvement of components of a graphical model while controlling potential new errors created in the process in other parts of the model. We demonstrate our methods in two physico-chemical examples, namely quantum scale-informed chemical kinetics and materials screening to improve the efficiency of fuel cells.

Probabilistic graphical models (PGM) are one of the rich frameworks. These models are used over complex domains for coding probability distributions. The joint distributions interact with each other over large numbers of random variables and are the combination of statistics and computer science. These concepts are dependent on theories such as probability theory, graph algorithms, machine learning, which make a basic tool in devising many machine learning problems. These are the origin for the contemporary methods in an extensive range of applications. These applications range as medical diagnosis, image understanding, speech recognition, natural language processing, etc. Graphical models are one of dominant tools for handling image processing applications. On the other hand, the volume of image data gives rise to a problem. The representation of all possible graphical model node variables with that of discrete states heads to the number of states for the model. This leads to interpretation computationally obstinate. Many projects involve a human intervention or an automated system to obtain the consensus established on existing information. The PGM, discussed in this chapter, offers a variety of approaches. The approach is based on models and allows interpretable models to be built which then is employed by reasoning algorithms. These models are also studied significantly from data and allow the approaches for cases where the model is manually built. Most real-world applications are of uncertain data which makes a model building more challenging. This chapter emphasizes on PGM where the uncertainty of data is obvious. PGM provides models that are more realistic. These are extended from Bayesian networks, undirected Markov networks, discrete and continuous models, and extensions to deal with dynamical systems and relational data also. For each class of models, the chapter describes the fundamental bases: representation, inference, and learning. Finally, the chapter considers the decision making under the uncertainty of the data.

Motion and structure estimation are elementary problems of computer vision. These are active areas of research, even though the first methods were proposed several decades ago. We develop new approaches for motion and structure estimation for autonomous driving. An autonomous vehicle requires an accurate model of its environment, wrong decisions made by an autonomous car can have severe consequences. We assume the monocular setup, where only a single camera is mounted on the car. Outdoor traffic sequences are challenging for optical flow estimation. The high speed of the car causes large displacements in the optical flow field, the lighting conditions are unstable and there can be strong distortions due to reflections and difficult weather conditions. We propose new discrete methods, which determine optical flow as optimal configuration of probabilistic graphical models. The first approach selects sparse locations in the reference frame, and matches them across the second image. The best correspondences, which match constraints from a multiple view configuration, are considered motion vectors in a graphical model. In a second approach, we solve for dense optical flow by approximating the original infeasible graphical model with a sequence of reduced models. The monocular configuration poses challenges to the estimation of scene structure, camera positions and scene parameters need to be estimated jointly. The geometry of multiple views creates blind spots in the images, and adds a scale ambiguity, which both to not exist in a setup with multiple cameras. We propose two methods for structure estimation. The first approach determines the energy optimal camera track, given optical flow and depth observations. A further approach estimates camera motion and a piecewise planar scene description jointly from a single optical flow field. The scene description contains depth and plane normal information. We evaluate our approaches for motion and structure estimation on different real world and rendered datasets. In addition to evaluation on publicly available evaluation data, we evaluate on a new rendered dataset with ground truth plane normals.

This thesis proposes a new parallel computing network--a binary relation inference network. The network topology is set up to match conventional optimization procedures. The network transforms some optimization problems (shortest path, transitive closure, assignment, etc.) into network convergence problems and divides the work load evenly among all processors. Link weights are known. It can operate in synchronous discrete-time, asynchronous discrete-time, and continuous-time domains. Inference network algorithms are derived directly from problem formulation, and are effective for a large range of problem sizes. The inference network consists of interconnected components; each component is composed of a unit, a number of sites attached to the unit, and links from the unit to sites on other units. Each unit is a simple computational engine. The topology of the inference network matches naturally to many optimization problems, since its deduction sites produce candidates to compete for the optimum and its decision-making units select the optimum from the site values. Either directly or through a transformation mapping to a systolic structure, discrete inference network algorithms can be implemented on a commercially available parallel processing facility, such as the Connection Machine. Analog inference network can be built using integrated circuits which solve problems efficiently in the continuous-time domain. In this thesis, mathematical analysis and numerical simulation were conducted for the synchronous, asynchronous, and analog inference network. Various applications has been discussed. The inference network has shown to solve the all-pair shortest path problem efficiently in various updating schemes. The inference network algorithm for the transitive closure problem parts derived straightforwardly from the formulation of the problem and the topology of the inference network. Efficient continuous-time solution was obtained from a non-synchronized logical circuit. The convergence of the network for the shortest path or transitive closure problems was shown to be independent of the problem size. It was demonstrated that the inference network can solve the assignment problem in a way similar to the Hopfield net's solution to the traveling salesman problem. However, the former was shown to converge for all test cases with problem size up to 128. (Abstract shortened by UMI.)

This paper proposes a new framework for image segmentation based on the integration of MRFs and deformable models using graphical models. We first construct a graphical model to represent the relationship of the observed image pixels, the true region labels and the underlying object contour. We then formulate the problem of image segmentation as the one of joint regioncontour inference and learning in the graphical model. The graphical model representation allows us to use an approximate structured variational inference technique to solve this otherwise intractable joint inference problem. Using this technique, the MAP solution to the original model is obtained by finding the MAP solutions of two simpler models, an extended MRF model and a probabilistic deformable model, iteratively and incrementally. In the extended MRF model, the true region labels are estimated using the BP algorithm in a band area around the estimated contour from the probabilistic deformable model, and the result in turn guides the probabilistic deformable model to an improved estimation of the contour. Experimental results show that our new hybrid method outperforms both the MRF-based and the deformable model-based methods.

Representing and processing lineages over uncertain data based on the Bayesian network

Robots had a great impact on the manufacturing industry ever since the early seventies when companies such as KUKA and ABB started deploying their first industrial robots. These robots merely performed very specific tasks in specific ways within well-defined environments. Still, they proved to be very useful as they could exceed human performance at these tasks. However, in order to enable robots to enter our daily life, they need to become more versatile and need to operate in much less structured environments. This thesis is partly devoted to stretching these limitations by means of learning, namely imitation learning (IL) and inverse reinforcement learning (IRL). Reinforcement learning (RL) is a powerful approach to enable robots to solve a task in an unknown environment. The practitioner describes a desired behavior by specifying a reward function and the robot autonomously interacts with the environment in order to find a control policy that generates high accumulated reward. However, RL is not suitable for teaching new tasks by non-experts because specifying appropriate reward functions can be difficult. Demonstrating the desired behavior is often easier for non-experts. Imitation learning can be used in order to enable the robot to reproduce the demonstrations. However, without explicitly inferring and modeling the intentions of the demonstrations, it can become difficult to solve the task for unseen situations. Inverse reinforcement learning (IRL) therefore aims to infer a reward function from the demonstrations, such that optimizing this reward function yields the desired behavior even for different situations. This thesis introduces a unifying approach to solve the inverse reinforcement learning problem in the same way as the reinforcement learning problem. This is achieved by framing both problems as information projection problems, i.e., we strive to minimize the relative entropy between a probabilistic model of the robot behavior and a given desired distribution. Furthermore, a trust region on the robot behavior is used to stabilize the optimization. For inverse reinforcement learning, the desired distribution is implicitly given by the expert demonstrations. The resulting optimization can be efficiently solved using state-of-the-art reinforcement learning methods. For reinforcement learning, the log-likelihood of the desired distribution is given by the reward function. The resulting optimization problem corresponds to a standard reinforcement learning formulation, except for an additional objective of maximizing the entropy of the robot behavior. This entropy objective adds little overhead to the optimization, but can lead to better exploration and more diversified policies. Trust-region I-projections are not only useful for training robots, but can also be applied to other machine learning problems. I-projections are typically used for variational inference, in order to approximate an intractable distribution by a simpler model. However, the resulting optimization problems are usually optimized based on stochastic gradient descent which often suffers from high variance in the gradient estimates. As trust-region I-projections where shown to be effective for reinforcement learning and inverse reinforcement learning, this thesis also explores their use for variational inference. More specifically, trust-region I-projections are investigated for the problem of approximating an intractable distribution by a Gaussian mixture model (GMM) with an adaptive number of components. GMMs are highly desirable for variational inference because they can yield arbitrary accurate approximations while inference from GMMs is still relatively cheap. In order to make learning the GMM feasible, we derive a lower bound that enables us to decompose the objective function. The optimization can then be performed by iteratively updating individual components using a technique from reinforcement learning. The resulting method is capable of learning approximations of significantly higher quality than existing variational inference methods. Due to the similarity of the underlying optimization problems, the insights gained from our variational inference method are also useful for IL and IRL. Namely, a similar lower bound can be applied also for the I-projection formulation of imitation learning. However, whereas for variational inference the lower bound serves to decompose the objective function, for imitation learning it allows us to provide a reward signal to the robot that does not depend on its behavior. Compared to reward functions that are relative to the current behavior of the robot---which are typical for popular adversarial methods---behavior-independent reward functions have the advantages that we can show convergence even for greedy optimization. Furthermore, behavior-independent reward functions solve the inverse reinforcement learning problem, thereby closing the gap between imitation learning and IRL. However, algorithms derived from our non-adversarial formulation are actually very similar to existing AIL methods, and we can even show that adversarial inverse reinforcement learning (AIRL) is indeed an instance of our formulation. AIRL was derived from an adversarial formulation, and we point out several problems of that derivation. In contrast, we show that AIRL can be straightforwardly derived from out non-adversarial formulation. Furthermore, we demonstrate that the non-adversarial formulation can be also used to derive novel algorithms by presenting a non-adversarial method for offline imitation learning.

Learning Markov boundaries from data without having to learn a Bayesian network first can be viewed as a feature subset selection problem and has received much attention due to its significance in ...