Semantic integration focuses on discovering, representing, and manipulating correspondences between entities in disparate data sources. The topic has been widely studied in the context of structured data, with problems being considered including ontology and schema matching, matching relational tuples, and reconciling inconsistent data values. In recent years, however, semantic integration over text has also received increasing attention. This article studies a key challenge in semantic integration over text: identifying whether different mentions of real-world entities, such as "JFK" and "John Kennedy," within and across natural language text documents, actually represent the same concept. We present a machine-learning study of this problem. The first approach is a discriminative approach -- a pairwise local classifier is trained in a supervised way to determine whether two given mentions represent the same real-world entity. This is followed, potentially, by a global clustering algorithm that uses the classifier as its similarity metric. Our second approach is a global generative model, at the heart of which is a view on how documents are generated and how names (of different entity types) are "sprinkled" into them. In its most general form, our model assumes (1) a joint distribution over entities (for example, a document that mentions "President Kennedy" is more likely to mention "Oswald" or "White House" than "Roger Clemens"), and (2) an "author" model that assumes that at least one mention of an entity in a document is easily identifiable and then generates other mentions via (3) an "appearance" model that governs how mentions are transformed from the "representative" mention. We show that both approaches perform very accurately, in the range of 90-95 percent. F1 measure for different entity types, much better than previous approaches to some aspects of this problem. Finally, we discuss how our solution for mention matching in text can be potentially applied to matching relational tuples, as well as to linking entities across databases and text.

There has been increased interest in devising learning techniques that combine unlabeled data with labeled data - i.e. semi-supervised learning. However, to the best of our knowledge, no study has been performed across various techniques and different types and amounts of labeled and unlabeled data. Moreover, most of the published work on semi-supervised learning techniques assumes that the labeled and unlabeled data come from the same distribution. It is possible for the labeling process to be associated with a selection bias such that the distributions of data points in the labeled and unlabeled sets are different. Not correcting for such bias can result in biased function approximation with potentially poor performance. In this paper, we present an empirical study of various semi-supervised learning techniques on a variety of datasets. We attempt to answer various questions such as the effect of independence or relevance amongst features, the effect of the size of the labeled and unlabeled sets and the effect of noise. We also investigate the impact of sample-selection bias on the semi -supervised learning techniques under study and implement a bivariate probit technique particularly designed to correct for such bias.

A general linear response method for deriving improved estimates of correlations in the variational Bayes framework is presented. Three applications are given and it is discussed how to use linear response as a general principle for improving mean field approximations.

We consider the question of how well a given distribution can be approximated with probabilistic graphical models. We introduce a new parameter, effective treewidth, that captures the degree of approximability as a tradeoff between the accuracy and the complexity of approximation. We present a simple approach to analyzing achievable tradeoffs that exploits the threshold behavior of monotone graph properties, and provide experimental results that support the approach.

We develop a framework based on Bayesian model averaging to explain how animals cope with uncertainty about contingencies in classical conditioning experiments. Traditional accounts of conditioning fit parameters within a fixed generative model of reinforcer delivery; uncertainty over the model structure is not considered. We apply the theory to explain the puzzling relationship between second-order conditioning and conditioned inhibition, two similar conditioning regimes that nonetheless result in strongly divergent behavioral outcomes. According to the theory, second-order conditioning results when limited experience leads animals to prefer a simpler world model that produces spurious correlations; conditioned inhibition results when a more complex model is justified by additional experience.

Recent work has examined the estimation of models of stimulus-driven neural activity in which some linear filtering process is followed by a nonlinear, probabilistic spiking stage. We analyze the estimation of one such model for which this nonlinear step is implemented by a noisy, leaky, integrate-and-fire mechanism with a spike-dependent aftercurrent. This model is a biophysically plausible alternative to models with Poisson (memory-less) spiking, and has been shown to effectively reproduce various spiking statistics of neurons in vivo. However, the problem of estimating the model from extracellular spike train data has not been examined in depth. We formulate the problem in terms of maximum likelihood estimation, and show that the computational problem of maximizing the likelihood is tractable.

When we learn a new motor skill, we have to contend with both the variability inherent in our sensors and the task. The sensory uncertainty can be reduced by using information about the distribution of previously experienced tasks. Here we impose a distribution on a novel sensorimotor task and manipulate the variability of the sensory feedback. We show that subjects internally represent both the distribution of the task as well as their sensory uncertainty. Moreover, they combine these two sources of information in a way that is qualitatively predicted by optimal Bayesian processing. We further analyze if the subjects can represent multimodal distributions such as mixtures of Gaussians. The results show that the CNS employs probabilistic models during sensorimotor learning even when the priors are multimodal.

We present a Bayesian approach to color constancy which utilizes a non-Gaussian probabilistic model of the image formation process. The parameters of this model are estimated directly from an uncalibrated image set and a small number of additional algorithmic parameters are chosen using cross validation. The algorithm is empirically shown to exhibit RMS error lower than other color constancy algorithms based on the Lambertian surface reflectance model when estimating the illuminants of a set of test images. This is demonstrated via a direct performance comparison utilizing a publicly available set of real world test images and code base.

Discrete Fourier transforms and other related Fourier methods have been practically implementable due to the fast Fourier transform (FFT). However there are many situations where doing fast Fourier transforms without complete data would be desirable. In this paper it is recognised that formulating the FFT algorithm as a belief network allows suitable priors to be set for the Fourier coefficients. Furthermore efficient generalised belief propagation methods between clusters of four nodes enable the Fourier coefficients to be inferred and the missing data to be estimated in near to O(n log n) time, where n is the total of the given and missing data points. This method is compared with a number of common approaches such as setting missing data to zero or to interpolation. It is tested on generated data and for a Fourier analysis of a damaged audio signal.

The E-step boils down to computing probabilities of the hidden variables given the observed variables (evidence) and current set of parameters. The M-step then, given these probabilities, yields a new set of parameters guaranteed to increase the likelihood. In Bayesian networks, that will be the focus of this article, the M-step is usually relatively straightforward. A complication may arise in the E-step, when computing the probability of the hidden variables given the evidence becomes intractable. An often used approach is to replace the exact yet intractable inference in the E step with approximate inference, either through sampling or using a deterministic variational method. The use of a "mean-field" variational method in this context leads to an algorithm known as variational EM and can be given theinterpretation of minimizing a free energy with respect to both a tractable approximate distribution (approximate E-step) and the parameters (M-step) [2]. Loopy belief propagation [3] and variants thereof, such as generalized belief propagation [4] and expectation propagation [5], have become popular alternatives to the "mean-field" variational approaches, often yielding somewhat better approximations. And indeed, they can and have been applied for approximate inference in the E-step of the EM algorithm (see e.g.