Methods for learning Bayesian network structure can discover dependency structure between observed variables, and have been shown to be useful in many applications. However, in domains that involve a large number of variables, the space of possible network structures is enormous, making it difficult, for both computational and statistical reasons, to identify a good model. In this paper, we consider a solution to this problem, suitable for domains where many variables have similar behavior. Our method is based on a new class of models, which we call module networks. A module network explicitly represents the notion of a module - a set of variables that have the same parents in the network and share the same conditional probability distribution. We define the semantics of module networks, and describe an algorithm that learns a module network from data. The algorithm learns both the partitioning of the variables into modules and the dependency structure between the variables. We evaluate our algorithm on synthetic data, and on real data in the domains of gene expression and the stock market. Our results show that module networks generalize better than Bayesian networks, and that the learned module network structure reveals regularities that are obscured in learned Bayesian networks.

Product models of low dimensional experts are a powerful way to avoid the curse of dimensionality. We present the ``under-complete product of experts' (UPoE), where each expert models a one dimensional projection of the data. The UPoE is fully tractable and may be interpreted as a parametric probabilistic model for projection pursuit. Its ML learning rules are identical to the approximate learning rules proposed before for under-complete ICA. We also derive an efficient sequential learning algorithm and discuss its relationship to projection pursuit density estimation and feature induction algorithms for additive random field models.

Bayesian networks are now being used in enormous fields, for example, diagnosis of a system, data mining, clustering and so on. In spite of their wide range of applications, the statistical properties have not yet been clarified, because the models are nonidentifiable and non-regular. In a Bayesian network, the set of its parameter for a smaller model is an analytic set with singularities in the space of large ones. Because of these singularities, the Fisher information matrices are not positive definite. In other words, the mathematical foundation for learning was not constructed. In recent years, however, we have developed a method to analyze non-regular models using algebraic geometry. This method revealed the relation between the models singularities and its statistical properties. In this paper, applying this method to Bayesian networks with latent variables, we clarify the order of the stochastic complexities.Our result claims that the upper bound of those is smaller than the dimension of the parameter space. This means that the Bayesian generalization error is also far smaller than that of regular model, and that Schwarzs model selection criterion BIC needs to be improved for Bayesian networks.

Conditional Random Fields (CRFs) are undirected graphical models, a special case of which correspond to conditionally-trained finite state machines. A key advantage of these models is their great flexibility to include a wide array of overlapping, multi-granularity, non-independent features of the input. In face of this freedom, an important question that remains is, what features should be used? This paper presents a feature induction method for CRFs. Founded on the principle of constructing only those feature conjunctions that significantly increase log-likelihood, the approach is based on that of Della Pietra et al [1997], but altered to work with conditional rather than joint probabilities, and with additional modifications for providing tractability specifically for a sequence model. In comparison with traditional approaches, automated feature induction offers both improved accuracy and more than an order of magnitude reduction in feature count; it enables the use of richer, higher-order Markov models, and offers more freedom to liberally guess about which atomic features may be relevant to a task. The induction method applies to linear-chain CRFs, as well as to more arbitrary CRF structures, also known as Relational Markov Networks [Taskar & Koller, 2002]. We present experimental results on a named entity extraction task.

Continuous time Bayesian networks (CTBNs) describe structured stochastic processes with finitely many states that evolve over continuous time. A CTBN is a directed (possibly cyclic) dependency graph over a set of variables, each of which represents a finite state continuous time Markov process whose transition model is a function of its parents. We address the problem of learning parameters and structure of a CTBN from fully observed data. We define a conjugate prior for CTBNs, and show how it can be used both for Bayesian parameter estimation and as the basis of a Bayesian score for structure learning. Because acyclicity is not a constraint in CTBNs, we can show that the structure learning problem is significantly easier, both in theory and in practice, than structure learning for dynamic Bayesian networks (DBNs). Furthermore, as CTBNs can tailor the parameters and dependency structure to the different time granularities of the evolution of different variables, they can provide a better fit to continuous-time processes than DBNs with a fixed time granularity.

We present and implement two algorithms for analytic asymptotic evaluation of the marginal likelihood of data given a Bayesian network with hidden nodes. As shown by previous work, this evaluation is particularly hard for latent Bayesian network models, namely networks that include hidden variables, where asymptotic approximation deviates from the standard BIC score. Our algorithms solve two central difficulties in asymptotic evaluation of marginal likelihood integrals, namely, evaluation of regular dimensionality drop for latent Bayesian network models and computation of non-standard approximation formulas for singular statistics for these models. The presented algorithms are implemented in Matlab and Maple and their usage is demonstrated for marginal likelihood approximations for Bayesian networks with hidden variables.

We consider the problem of learning a Riemannian metric associated with a given differentiable manifold and a set of points. Our approach to the problem involves choosing a metric from a parametric family that is based on maximizing the inverse volume of a given dataset of points. From a statistical perspective, it is related to maximum likelihood under a model that assigns probabilities inversely proportional to the Riemannian volume element. We discuss in detail learning a metric on the multinomial simplex where the metric candidates are pullback metrics of the Fisher information under a continuous group of transformations. When applied to documents, the resulting geodesics resemble, but outperform, the TFIDF cosine similarity measure in classification.

Graphical models with bi-directed edges (<->) represent marginal independence: the absence of an edge between two vertices indicates that the corresponding variables are marginally independent. In this paper, we consider maximum likelihood estimation in the case of continuous variables with a Gaussian joint distribution, sometimes termed a covariance graph model. We present a new fitting algorithm which exploits standard regression techniques and establish its convergence properties. Moreover, we contrast our procedure to existing estimation methods.

Learning with hidden variables is a central challenge in probabilistic graphical models that has important implications for many real-life problems. The classical approach is using the Expectation Maximization (EM) algorithm. This algorithm, however, can get trapped in local maxima. In this paper we explore a new approach that is based on the Information Bottleneck principle. In this approach, we view the learning problem as a tradeoff between two information theoretic objectives. The first is to make the hidden variables uninformative about the identity of specific instances. The second is to make the hidden variables informative about the observed attributes. By exploring different tradeoffs between these two objectives, we can gradually converge on a high-scoring solution. As we show, the resulting, Information Bottleneck Expectation Maximization (IB-EM) algorithm, manages to find solutions that are superior to standard EM methods.

The Hierarchical Mixture of Experts (HME) is a well-known tree-based model for regression and classification, based on soft probabilistic splits. In its original formulation it was trained by maximum likelihood, and is therefore prone to over-fitting. Furthermore the maximum likelihood framework offers no natural metric for optimizing the complexity and structure of the tree. Previous attempts to provide a Bayesian treatment of the HME model have relied either on ad-hoc local Gaussian approximations or have dealt with related models representing the joint distribution of both input and output variables. In this paper we describe a fully Bayesian treatment of the HME model based on variational inference. By combining local and global variational methods we obtain a rigourous lower bound on the marginal probability of the data under the model. This bound is optimized during the training phase, and its resulting value can be used for model order selection. We present results using this approach for a data set describing robot arm kinematics.