Deep generative models parameterized by neural networks have recently achieved state-ofthe-art performance in unsupervised and semisupervised learning. We extend deep generative models with auxiliary variables which improves the variational approximation. The auxiliary variables leave the generative model unchanged but make the variational distribution more expressive. Inspired by the structure...

January is is the second progress report for Project /// (Generative Kernels and Score Spaces for Classiication of Speech) within the Global Uncertainties Programme. is project combines the current generative models developed in the speech community with discriminative classiiers. An important aspect of the approach is that the generative models are used to deene a score-space that can be used ...

Generative tools for programming language support have a long history. Tools for interactive language-aware software development are central to the HARMONIA object-oriented framework. However, the generative aspects of HARMONIA are implemented in an ad hoc fashion. This paper explains how systematic generative programming could be used to improve the implementation of HARMONIA and similar systems.

Generative modeling of high dimensional data like images is a notoriously difficult and ill-defined problem. In particular, how to evaluate a learned generative model is unclear. In this paper, we argue that adversarial learning, pioneered with generative adversarial networks (GANs), provides an interesting framework to implicitly define more meaningful task losses for unsupervised tasks, such ...

GENERATIVE CAPACITY was introduced by Chomsky (1963) in the context of the theory of formal grammars and automata (→ Finite State Grammars and Languages, → Context Free Grammars and Languages, → Mildly Context Sensitive Grammars and Languages, → Automata Theory). A language is defined as a set of strings over some vocabulary (e.g., a set of sentences over a vocabulary of words). A formal gramma...

This article lays out an approach that combines a formal-generative perspective on language, including tolerance of abstract analyses, with a typological focus on comparing unrelated languages from around the world. It argues that this can be a powerful combination for discovering linguistic universals and patterns in linguistic variation that are not detected by other means.

Generative score spaces provide a principled method to exploit generative information, e.g., data distribution and hidden variables, in discriminative classifiers. The underlying methodology is to derive measures or score functions from generative models. The derived score functions, spanning the so-called score space, provide features of a fixed dimension for discriminative classification. In ...

This article presents a functional model of learning from teaching that, in contrast to structural models of schemata and knowledge representation, focuses on the neural and cognitive processes that learners use to generate meaning and understanding from instruction. Wittrock's model of generative learning (Wittrock, 1974a, 1990) consists of four major processes: (a) attention, (b) motivation, ...

The field of procedural content generation continues to grow in scope and in technology, but the term “procedural content generation” awkwardly suggests that the field’s output be defined by its ability to produce game “content”, a term that fails to capture the breadth of artifacts produced by PCG researchers. There are many parallel fields of research on using algorithmic means to generate wh...