Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models

2020/02/17 by Chin-Wei Huang, Huang, Chin-Wei, Laurent Dinh +4 · 6 citations
Computer Science · Engineering · Mathematics · #3D Shape Modeling and Analysis #Computer Graphics and Visualization Techniques #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Human Motion and Animation #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.2002.07101

27 pages, 12 figures

arxiv created 2020/02/17 · openalex publication_date 2020/02/17 · arxiv updated 2020/02/19 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28

Abstract

In this work, we propose a new family of generative flows on an augmented data space, with an aim to improve expressivity without drastically increasing the computational cost of sampling and evaluation of a lower bound on the likelihood. Theoretically, we prove the proposed flow can approximate a Hamiltonian ODE as a universal transport map. Empirically, we demonstrate state-of-the-art performance on standard benchmarks of flow-based generative modeling.

Citations

• A Recurrent Latent Variable Model for Sequential Data
• NICE: Non-linear Independent Components Estimation
• Auto-Encoding Variational Bayes
• Gradient-based learning applied to document recognition
• Pixel Recurrent Neural Networks
• Generative Adversarial Networks as Variational Training of Energy Based Models
• Variational Lossy Autoencoder
• PixelVAE: A Latent Variable Model for Natural Images
• Multiplicative Normalizing Flows for Variational Bayesian Neural Networks
• Calibrating Energy-based Generative Adversarial Networks
• Learnable Explicit Density for Continuous Latent Space and Variational Inference
• Hierarchical Adversarially Learned Inference
• Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design
• Semi-Conditional Normalizing Flows for Semi-Supervised Learning
• Improving Exploration in Soft-Actor-Critic with Normalizing Flows Policies
• Stochastic Neural Network with Kronecker Flow
• Variational Inference with Normalizing Flows
• Conditional Image Generation with PixelCNN Decoders
• FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models
• Relaxing Bijectivity Constraints with Continuously Indexed Normalising Flows
• Learning Multiple Layers of Features from Tiny Images

Cited by

• Fine-tuning Flow Matching Generative Models with Intermediate Feedback
• Scalable Boltzmann Generators for equilibrium sampling of large-scale materials
• VFP: Variational Flow-Matching Policy for Multi-Modal Robot Manipulation
• Ellipsoidal Time Series Forecasting
• Accelerating Continuous Normalizing Flow with Trajectory Polynomial Regularization
• NVAE: A Deep Hierarchical Variational Autoencoder

Related

• The Expressive Power of a Class of Normalizing Flow Models
• Variations and Relaxations of Normalizing Flows
• Jet: A Modern Transformer-Based Normalizing Flow
• Free-form Flows: Make Any Architecture a Normalizing Flow
• Normalizing Flows are Capable Generative Models