Transformation Autoregressive Networks

TL;DR

This paper systematically characterizes density estimation methods, introduces novel models combining autoregressive and transformation approaches, and demonstrates their effectiveness in tasks like outlier detection and image modeling.

Contribution

It proposes new methods that integrate autoregressive models with variable transformations, advancing density estimation techniques.

Findings

Jointly leveraging transformations and autoregressive models improves performance.

The models are effective in outlier detection and image modeling.

A new data-driven framework for learning distribution families is introduced.

Abstract

The fundamental task of general density estimation $p(x)$ has been of keen interest to machine learning. In this work, we attempt to systematically characterize methods for density estimation. Broadly speaking, most of the existing methods can be categorized into either using: \textit{a}) autoregressive models to estimate the conditional factors of the chain rule, $p(x_{i}\, |\, x_{i-1}, \ldots)$; or \textit{b}) non-linear transformations of variables of a simple base distribution. Based on the study of the characteristics of these categories, we propose multiple novel methods for each category. For example we proposed RNN based transformations to model non-Markovian dependencies. Further, through a comprehensive study over both real world and synthetic data, we show for that jointly leveraging transformations of variables and autoregressive conditional models, results in a considerable improvement in performance. We illustrate the use of our models in outlier detection and image modeling. Finally we introduce a novel data driven framework for learning a family of distributions.