Bidirectional Generative Modeling Using Adversarial Gradient Estimation

Xinwei Shen, Tong Zhang, Kani Chen
2/21/2020
cs.LGcs.CVstat.ML

Official code: xwshen51/AGES (6 stars, last commit 2024-07-25).

Abstract

This paper considers the general $f$-divergence formulation of bidirectional generative modeling, which includes VAE and BiGAN as special cases. We present a new optimization method for this formulation, where the gradient is computed using an adversarially learned discriminator. In our framework, we show that different divergences induce similar algorithms in terms of gradient evaluation, except with different scaling. Therefore this paper gives a general recipe for a class of principled $f$-divergence based generative modeling methods. Theoretical justifications and extensive empirical studies are provided to demonstrate the advantage of our approach over existing methods.

View on arXivView PDF

Code Implementations(1)

xwshen51/AGESOfficial100%

Bidirectional Generative Modeling Using Adversarial Gradient Estimation

62Shell, PythonFeb 21, 20202 years ago

Similar Papers

Seasoning Generative Models for a Generalization Aftertaste

86% match2026

Deep Generative Learning via Variational Gradient Flow

83% match1 repo122019

f-Divergence Based Classification: Beyond the Use of Cross-Entropy

82% match2 repos62024

Robust Generative Learning with Lipschitz-Regularized α-Divergences Allows Minimal Assumptions on Target Distributions

81% match2024

Precision-Recall Divergence Optimization for Generative Modeling with GANs and Normalizing Flows

81% match2023

Aligning Language Models with Preferences through f-divergence Minimization

81% match1 repo752023

Cite this paper

@article{shen2020bidirectional,
  title  = {Bidirectional Generative Modeling Using Adversarial Gradient Estimation},
  author = {Xinwei Shen and Tong Zhang and Kani Chen},
  year   = {2020},
  eprint = {2002.09161},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2002.09161}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2002.09161.svg)](https://codeofpaper.com/paper/2002.09161)

Discussion