The continuous time Bayesian network (CTBN) is a probabilistic graphical model that enables reasoning about complex, interdependent, and continuous-time subsystems. The model uses nodes to denote subsystems and arcs to denote conditional dependence. This dependence manifests in how the dynamics of a subsystem change based on the current states of its parents in the network. While the original CTBN definition allows users to specify the dynamics of how the system evolves, users might also want to place value expressions over the dynamics of the model in the form of performance functions. We formalize these performance functions for the CTBN and show how they can be factored in the same way as the network, allowing what we argue is a more intuitive and explicit representation. For cases in which a performance function must involve multiple nodes, we show how to augment the structure of the CTBN to account for the performance interaction while maintaining the factorization of a single performance function for each node.

Endowing the semantically-oblivious Internet with Intelligence would advance the Internet capability to learn traffic behavior and to predict future events. In this paper, we propose a hybrid intelligence memory system, or NetMem, for network-semantics reasoning and targeting Internet intelligence. NetMem provides a memory structure, mimicking the human memory functionalities, via short-term memory (StM) and long-term memory (LtM). NetMem has the capability to build runtime accessible dynamic network-concept ontology (DNCO) at different levels of granularity. We integrate Latent Dirichlet Allocation (LDA) and Hidden Markov Models (HMM) to extract network-semantics based on learning patterns and recognizing features with syntax and semantic dependencies. Due to the large scale and high-dimensionality of Internet data, we utilize the Locality Sensitive Hashing (LSH) algorithm for data dimensionality reduction. Simulation results using real network traffic show that NetMem with hybrid intelligence learn traffic data semantics effectively and efficiently even with significant reduction in volume and dimensionality of data, thus enhancing Internet intelligence for self-/situation-awareness and event/behavior prediction.

Robots may encounter undesirable outcomes due to failures during the execution of their plans in the physical world. Failures should be detected, and the underlying reasons should be found by the robot in order to handle these failure situations efficiently. Sometimes, there may be more than one cause of a failure, and they are not necessarily related to the action in execution. In this paper, we propose a temporal and Hierarchical Hidden Markov Model (HHMM) based failure isolation method. These HHMMs run in parallel to determine causes of unexpected deviations. Experiments on our Pioneer 3-AT robot show that our method successfully isolates failures suggesting possible causes.

We report the results of an empirical evaluation of structural simplification of Bayesian networks by removing weak arcs. We conduct a series of experiments on six networks built from real data sets selected from the UC Irvine Machine Learning Repository. We systematically remove arcs from the weakest to the strongest, relying on four measures of arc strength, and measure the classification accuracy of the resulting simplified models. Our results show that removing up to roughly 20 percent of the weakest arcs in a network has minimal effect on its classification accuracy. At the same time, structural simplification of networks leads to significant reduction of both the amount of memory taken by the clique tree and the amount of computation needed to perform inference.

A new approach to maximum likelihood learning of discrete graphical models and RBM in particular is introduced. Our method, Perturb and Descend (PD) is inspired by two ideas (I) perturb and MAP method for sampling (II) learning by Contrastive Divergence minimization. In contrast to perturb and MAP, PD leverages training data to learn the models that do not allow efficient MAP estimation. During the learning, to produce a sample from the current model, we start from a training data and descend in the energy landscape of the "perturbed model", for a fixed number of steps, or until a local optima is reached. For RBM, this involves linear calculations and thresholding which can be very fast. Furthermore we show that the amount of perturbation is closely related to the temperature parameter and it can regularize the model by producing robust features resulting in sparse hidden layer activation.

We present a machine learning framework for modeling protein dynamics. Our approach uses L1-regularized, reversible hidden Markov models to understand large protein datasets generated via molecular dynamics simulations. Our model is motivated by three design principles: (1) the requirement of massive scalability; (2) the need to adhere to relevant physical law; and (3) the necessity of providing accessible interpretations, critical for both cellular biology and rational drug design. We present an EM algorithm for learning and introduce a model selection criteria based on the physical notion of convergence in relaxation timescales. We contrast our model with standard methods in biophysics and demonstrate improved robustness. We implement our algorithm on GPUs and apply the method to two large protein simulation datasets generated respectively on the NCSA Bluewaters supercomputer and the Folding@Home distributed computing network. Our analysis identifies the conformational dynamics of the ubiquitin protein critical to cellular signaling, and elucidates the stepwise activation mechanism of the c-Src kinase protein.

Contrastive divergence (CD) is a promising method of inference in high dimensional distributions with intractable normalizing constants, however, the theoretical foundations justifying its use are somewhat shaky. This document proposes a framework for understanding CD inference, how/when it works, and provides multiple justifications for the CD moment conditions, including framing them as a variational approximation. Algorithms for performing inference are discussed and are applied to social network data using an exponential-family random graph models (ERGM). The framework also provides guidance about how to construct MCMC kernels providing good CD inference, which turn out to be quite different from those used typically to provide fast global mixing.

Developing feature selection algorithms that move beyond a pure correlational to a more causal analysis of observational data is an important problem in the sciences. Several algorithms attempt to do so by discovering the Markov blanket of a target, but they all contain a forward selection step which variables must pass in order to be included in the conditioning set. As a result, these algorithms may not consider all possible conditional multivariate combinations. We improve on this limitation by proposing a backward elimination method that uses a kernel-based conditional dependence measure to identify the Markov blanket in a fully multivariate fashion. The algorithm is easy to implement and compares favorably to other methods on synthetic and real datasets.

This paper proposes an online tree-based Bayesian approach for reinforcement learning. For inference, we employ a generalised context tree model. This defines a distribution on multivariate Gaussian piecewise-linear models, which can be updated in closed form. The tree structure itself is constructed using the cover tree method, which remains efficient in high dimensional spaces. We combine the model with Thompson sampling and approximate dynamic programming to obtain effective exploration policies in unknown environments. The flexibility and computational simplicity of the model render it suitable for many reinforcement learning problems in continuous state spaces. We demonstrate this in an experimental comparison with least squares policy iteration.

A sequence of random variables is exchangeable if its joint distribution is invariant under variable permutations. We introduce exchangeable variable models (EVMs) as a novel class of probabilistic models whose basic building blocks are partially exchangeable sequences, a generalization of exchangeable sequences. We prove that a family of tractable EVMs is optimal under zero-one loss for a large class of functions, including parity and threshold functions, and strictly subsumes existing tractable independence-based model families. Extensive experiments show that EVMs outperform state of the art classifiers such as SVMs and probabilistic models which are solely based on independence assumptions.