The quality of data representation in deep learning methods is directly related to the prior model imposed on the representations; however, generally used fixed priors are not capable of adjusting to the context in the data. To address this issue, we propose deep predictive coding networks, a hierarchical generative model that empirically alters priors on the latent representations in a dynamic and context-sensitive manner. This model captures the temporal dependencies in time-varying signals and uses top-down information to modulate the representation in lower layers. The centerpiece of our model is a novel procedure to infer sparse states of a dynamic model which is used for feature extraction. We also extend this feature extraction block to introduce a pooling function that captures locally invariant representations. When applied on a natural video data, we show that our method is able to learn high-level visual features. We also demonstrate the role of the top-down connections by showing the robustness of the proposed model to structured noise.

Relevance-based explanation is a scheme in which partial assignments to Bayesian belief network variables are explanations (abductive conclusions). We allow variables to remain unassigned in explanations as long as they are irrelevant to the explanation, where irrelevance is defined in terms of statistical independence. When multiple-valued variables exist in the system, especially when subsets of values correspond to natural types of events, the overspecification problem, alleviated by independence-based explanation, resurfaces. As a solution to that, as well as for addressing the question of explanation specificity, it is desirable to collapse such a subset of values into a single value on the fly. The equivalent method, which is adopted here, is to generalize the notion of assignments to allow disjunctive assignments. We proceed to define generalized independence based explanations as maximum posterior probability independence based generalized assignments (GIB-MAPs). GIB assignments are shown to have certain properties that ease the deJ ign of algorithms for computing GIB-MAPs. One such algorithm is discussed here, as well as suggestions for how other algorithms may be adapted to compute GIB-MAPs. GIB-MAP explanations still suffer from instability, a problem which may be addressed using "approximate" conditional independence as a condition for irrelevance.

Mauricio Sadinle is a Ph.D. student, Department of Statistics, Carnegie Mellon University, Pittsburgh, PA 15213 (email: msadinle@stat.cmu.edu); and Stephen E. Fienberg is Maurice Falk University Professor of Statistics and Social Science in the Department of Statistics, the Machine Learning Department, and the Heinz College, Carnegie Mellon University (email: fien-berg@stat.cmu.edu). This research was partially supported by NSF Grants BCS-0941518 and SES-1130706 to Carnegie Mellon University, and by the Singapore National Research Foundation under its International Research Centre @ Singapore Funding Initiative and administered by the IDM Programme Office. The authors thank Rob Hall, Kristian Lum, Michael Larsen, the Associate Editor and two referees for helpful comments and suggestions on earlier versions of this paper, and Jorge A. Restrepo for providing the Colombian homicide data. An early version of this paper was written by the first author when he was affiliated to the Conflict Analysis Resource Center (CERAC) and the National University of Colombia at Bogot a. Abstract We present a probabilistic method for linking multiple datafiles. This task is not trivial in the absence of unique identifiers for the individuals recorded. This is a common scenario when linking census data to coverage measurement surveys for census coverage evaluation, and in general when multiple record-systems need to be integrated for posterior analysis. The goal of multiple record linkage is to classify the recordK -tuples coming fromK datafiles according to the different matching patterns. We use a mixture model to fit matching probabilities via maximum likelihood using the EM algorithm. We present a method to decide the recordK -tuples membership to the subsets of matching patterns and we prove its optimality. We apply our method to the integration of the three Colombian homicide record systems and perform a simulation study to explore the performance of the method under measurement error and different scenarios. The proposed method works well and opens new directions for future research. Key words and phrases: Bell number; Census undercount; Data linkage; Data matching; EM algorithm; Mixture model; Multiple systems estimation; Partially ordered set. 1 INTRODUCTION Record linkage is a widely-used technique for identifying records that refer to the same individual across different datafiles. This task is not trivial when unique identifiers are not available, and many authors have proposed probabilistic methods to deal with this problem building upon the seminal work of Newcombe et al. (1959) and Fellegi and Sunter (1969).

Distributed knowledge based applications in open domain rely on common sense information which is bound to be uncertain and incomplete. To draw the useful conclusions from ambiguous data, one must address uncertainties and conflicts incurred in a holistic view. No integrated frameworks are viable without an in-depth analysis of conflicts incurred by uncertainties. In this paper, we give such an analysis and based on the result, propose an integrated framework. Our framework extends definite argumentation theory to model uncertainty. It supports three views over conflicting and uncertain knowledge. Thus, knowledge engineers can draw different conclusions depending on the application context (i.e. view). We also give an illustrative example on strategical decision support to show the practical usefulness of our framework.

The deontic logic DUS is a Deontic Update Semantics for prescriptive obligations based on the update semantics of Veltman. In DUS the definition of logical validity of obligations is not based on static truth values but on dynamic action transitions. In this paper prescriptive defeasible obligations are formalized in update semantics and the diagnostic problem of defeasible deontic logic is discussed. Assume a defeasible obligation `normally A ought to be (done)' together withthe fact `A is not (done).' Is this an exception of the normality claim, or is it a violation of the obligation? In this paper we formalize the heuristic principle that it is a violation, unless there is a more specific overriding obligation. The underlying motivation from legal reasoning is that criminals should have as little opportunities as possible to excuse themselves by claiming that their behavior was exceptional rather than criminal.

Inference networks have a variety of important uses and are constructed by persons having quite different standpoints. Discussed in this paper are three different but complementary methods for generating and analyzing probabilistic inference networks. The first method, though over eighty years old, is very useful for knowledge representation in the task of constructing probabilistic arguments. It is also useful as a heuristic device in generating new forms of evidence. The other two methods are formally equivalent ways for combining probabilities in the analysis of inference networks. The use of these three methods is illustrated in an analysis of a mass of evidence in a celebrated American law case.

The problem of topic modeling can be seen as a generalization of the clustering problem, in that it posits that observations are generated due to multiple latent factors (e.g., the words in each document are generated as a mixture of several active topics, as opposed to just one). This increased representational power comes at the cost of a more challenging unsupervised learning problem of estimating the topic probability vectors (the distributions over words for each topic), when only the words are observed and the corresponding topics are hidden. We provide a simple and efficient learning procedure that is guaranteed to recover the parameters for a wide class of mixture models, including the popular latent Dirichlet allocation (LDA) model. For LDA, the procedure correctly recovers both the topic probability vectors and the prior over the topics, using only trigram statistics (i.e., third order moments, which may be estimated with documents containing just three words). The method, termed Excess Correlation Analysis (ECA), is based on a spectral decomposition of low order moments (third and fourth order) via two singular value decompositions (SVDs). Moreover, the algorithm is scalable since the SVD operations are carried out on $k\times k$ matrices, where $k$ is the number of latent factors (e.g. the number of topics), rather than in the $d$-dimensional observed space (typically $d \gg k$).

In many medical studies, estimating the failure time distribution function, or quantities that depend on this distribution, as a function of patient demographic and prognostic variables, is of central importance for risk assessment and health planing. Frequently, such data is subject to right censoring. The goal of this paper is to develop tools for analyzing such data using machine learning techniques. Traditional approaches to right censored failure time analysis include using parametric models, such as the Weibull distribution, and semiparametric models such as proportional hazard models (see Lawless, 2003, for both). Even when less stringent models--such as nonparametric estimation--are used, it is typically assumed that the distribution function is smooth in both time and covariates (Dabrowska, 1987; Gonzalez-Manteiga and Cadarso-Suarez, 1994). These assumptions seem restrictive, especially when considering today's high-dimensional data settings.

Determinantal point processes (DPPs) are elegant probabilistic models of repulsion that arise in quantum physics and random matrix theory. In contrast to traditional structured models like Markov random fields, which become intractable and hard to approximate in the presence of negative correlations, DPPs offer efficient and exact algorithms for sampling, marginalization, conditioning, and other inference tasks. We provide a gentle introduction to DPPs, focusing on the intuitions, algorithms, and extensions that are most relevant to the machine learning community, and show how DPPs can be applied to real-world applications like finding diverse sets of high-quality search results, building informative summaries by selecting diverse sentences from documents, modeling non-overlapping human poses in images or video, and automatically building timelines of important news stories.

The direct effect of one event on another can be defined and measured by holding constant all intermediate variables between the two. Indirect effects present conceptual and practical difficulties (in nonlinear models), because they cannot be isolated by holding certain variables constant. This paper presents a new way of defining the effect transmitted through a restricted set of paths, without controlling variables on the remaining paths. This permits the assessment of a more natural type of direct and indirect effects, one that is applicable in both linear and nonlinear models and that has broader policy-related interpretations. The paper establishes conditions under which such assessments can be estimated consistently from experimental and nonexperimental data, and thus extends path-analytic techniques to nonlinear and nonparametric models.