We propose a Topic Compositional Neural Language Model (TCNLM), a novel method designed to simultaneously capture both the global semantic meaning and the local wordordering structure in a document. The TCNLM learns the global semantic coherence of a document via a neural topic model, and the probability of each learned latent topic is further used to build a Mixture-ofExperts (MoE) language mo...

This analysis joins the debate on how declines in marriage have shifted the composition of the unmarried and married populations in the United States, and how compositional shifts have affected nonmarital birth rates. Gray, Stockard, and Stone (2006) presented one model for compositional effects that Ermisch (2009) challenged with alternative statistical tests. I propose an alternative model fo...

The Continuous Stochastic Logic (CSL) is a powerful means to state properties which refer to Continuous Time Markov Chains (CTMCs). The verification of such properties on a model can be achieved through a suitable algorithm. In this doctoral thesis, the CSL logic has been considered and two major aspects have been addressed: the analysis of its expressiveness and the study of methods for a deco...

We often build complex probabilistic models by composing simpler models—using one model to generate parameters or latent variables for another model. This allows us to express complex distributions over the observed data and to share statistical structure between different parts of a model. In this thesis, we present a space of matrix decomposition models defined by the composition of a small n...

We develop a model to relate a multivariate compositional response to a number of covariates. We propose a new graphical model, called the Random Effects Discrete Regression (REDR) model, which allows for examination of the complex conditional relationships between a set of covariates and multiple discrete response variables. Our approach offers a number of advantages over previous approaches a...

Using kernel density estimation and mixture models, household size-adjusted income distributions in Italy are cross-sectionally examined over the period 1987-2000. Nonparametric tests assess the shape time invariance and the presence of modes in the distributions. Evidence shows that income tend to cluster around more than one point, giving good reasons to model the shapes by a finite mixture d...

Compositional data need a special treatment prior to correlation analysis. In this paper we argue why standard transformations for compositional data are not suitable for computing correlations, and why the use of raw or log-transformed data is neither meaningful. As a solution, a procedure based on balances is outlined, leading to sensible correlation measures. The construction of the balances...

This dissertation presents a framework for the application of compositional modularity a module model that facilitates extensive reuse of highly decomposed software Compositional modularity supports not only the traditional notions of program decomposition and encapsulation but also e ective mechanisms for module recom position Based on a previously developed model a suite of operators individu...

For the past four decades the compositional organization of the mammalian genome posed a formidable challenge to molecular evolutionists attempting to explain it from an evolutionary perspective. Unfortunately, most of the explanations adhered to the "isochore theory," which has long been rebutted. Recently, an alternative compositional domain model was proposed depicting the human and cow geno...

Compositional embedding models build a representation (or embedding) for a linguistic structure based on its component word embeddings. We propose a Feature-rich Compositional Embedding Model (FCM) for relation extraction that is expressive, generalizes to new domains, and is easy-to-implement. The key idea is to combine both (unlexicalized) handcrafted features with learned word embeddings. Th...