Transformation Autoregressive Networks

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, $p(x_i|x_{i-1},\dots )$; 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.

References in corpus (5)

• NICE: Non-linear Independent Components Estimation
• Modeling Temporal Dependencies in High-Dimensional Sequences: Application to Polyphonic Music Generation and Transcription
• MADE: Masked Autoencoder for Distribution Estimation
• Masked Autoregressive Flow for Density Estimation
• Contextual Explanation Networks

Cited by in corpus (10)

• FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models
• FloWaveNet : A Generative Flow for Raw Audio
• TraDE: Transformers for Density Estimation
• Flow Models for Arbitrary Conditional Likelihoods
• Neural Density Estimation and Likelihood-free Inference
• Gaussianization Flows
• Improving Sequential Latent Variable Models with Autoregressive Flows
• A Forest from the Trees: Generation through Neighborhoods
• Arbitrary Conditional Distributions with Energy
• Accelerating Continuous Normalizing Flow with Trajectory Polynomial Regularization