Towards Principled Methods for Training Generative Adversarial Networks

Abstract

The goal of this paper is not to introduce a single algorithm or method, but to make theoretical steps towards fully understanding the training dynamics of generative adversarial networks. In order to substantiate our theoretical analysis, we perform targeted experiments to verify our assumptions, illustrate our claims, and quantify the phenomena. This paper is divided into three sections. The first section introduces the problem at hand. The second section is dedicated to studying and proving rigorously the problems including instability and saturation that arize when training generative adversarial networks. The third section examines a practical and theoretically grounded direction towards solving these problems, while introducing new tools to study them.

References in corpus (5)

• Explaining and Harnessing Adversarial Examples
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
• Improved Techniques for Training GANs

Cited by in corpus (106)

• The relativistic discriminator: a key element missing from standard GAN
• Data Synthesis based on Generative Adversarial Networks
• Generative Adversarial Networks in Computer Vision: A Survey and Taxonomy
• DARTS+: Improved Differentiable Architecture Search with Early Stopping
• MD-GAN: Multi-Discriminator Generative Adversarial Networks for Distributed Datasets
• Towards the Automatic Anime Characters Creation with Generative Adversarial Networks
• Unsupervised Domain Adaptation with Adversarial Residual Transform Networks
• ChemGAN challenge for drug discovery: can AI reproduce natural chemical diversity?
• Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step
• A review of Generative Adversarial Networks for Electronic Health Records: applications, evaluation measures and data sources
• Super-resolution emulator of cosmological simulations using deep physical models
• Model-based Deep Reinforcement Learning for Dynamic Portfolio Optimization
• On denoising autoencoders trained to minimise binary cross-entropy
• Parametrization and generation of geological models with generative adversarial networks
• A Systematic Survey of Regularization and Normalization in GANs
• A Latent Encoder Coupled Generative Adversarial Network (LE-GAN) for Efficient Hyperspectral Image Super-resolution
• Implicit Maximum Likelihood Estimation
• Spatial Evolutionary Generative Adversarial Networks
• Improving the Improved Training of Wasserstein GANs: A Consistency Term and Its Dual Effect
• Tutorial: Deriving the Standard Variational Autoencoder (VAE) Loss Function
• Unsupervised Text Style Transfer using Language Models as Discriminators
• AutoGAN: Neural Architecture Search for Generative Adversarial Networks
• Convergence Problems with Generative Adversarial Networks (GANs)
• Comparison of Maximum Likelihood and GAN-based training of Real NVPs
• A Convex Duality Framework for GANs
• COT-GAN: Generating Sequential Data via Causal Optimal Transport
• Wasserstein Generative Adversarial Uncertainty Quantification in Physics-Informed Neural Networks
• Face Transfer with Generative Adversarial Network
• Safer Classification by Synthesis
• MGGAN: Solving Mode Collapse using Manifold Guided Training
• How Well Generative Adversarial Networks Learn Distributions
• MTS-DVGAN: Anomaly Detection in Cyber-Physical Systems using a Dual Variational Generative Adversarial Network
• Laplacian Smoothing Gradient Descent
• Improved Training of Generative Adversarial Networks Using Representative Features
• CnGAN: Generative Adversarial Networks for Cross-network user preference generation for non-overlapped users
• From EMBER to FIRE: predicting high resolution baryon fields from dark matter simulations with Deep Learning
• On the capacity of deep generative networks for approximating distributions
• Machine Learning Methods Economists Should Know About
• Neural Language Generation: Formulation, Methods, and Evaluation
• Inverting The Generator Of A Generative Adversarial Network (II)
• High-Quality Face Image SR Using Conditional Generative Adversarial Networks
• SARM: Sparse Autoregressive Model for Scalable Generation of Sparse Images in Particle Physics
• CoT: Cooperative Training for Generative Modeling of Discrete Data
• A unified framework for 21cm tomography sample generation and parameter inference with Progressively Growing GANs
• Adversarial Variational Optimization of Non-Differentiable Simulators
• eCommerceGAN : A Generative Adversarial Network for E-commerce
• Relaxed Wasserstein with Applications to GANs
• Lipschitzness Is All You Need To Tame Off-policy Generative Adversarial Imitation Learning
• Generative Adversarial Networks (GANs): What it can generate and What it cannot?
• Learning stochastic object models from medical imaging measurements by use of advanced ambient generative adversarial networks
• On the Implicit Assumptions of GANs
• Quantitatively Evaluating GANs With Divergences Proposed for Training
• Lipizzaner: A System That Scales Robust Generative Adversarial Network Training
• Training Generative Adversarial Networks by Solving Ordinary Differential Equations
• Comparing Generative Adversarial Network Techniques for Image Creation and Modification
• Fluid Flow Mass Transport for Generative Networks
• Non-parametric estimation of Jensen-Shannon Divergence in Generative Adversarial Network training
• Variational f-divergence Minimization
• Enforcing constraints for time series prediction in supervised, unsupervised and reinforcement learning
• An Improved Self-supervised GAN via Adversarial Training
• Training Generative Adversarial Networks via Primal-Dual Subgradient Methods: A Lagrangian Perspective on GAN
• LGAN: Lung Segmentation in CT Scans Using Generative Adversarial Network
• Accurate generation of stochastic dynamics based on multi-model Generative Adversarial Networks
• MR-GAN: Manifold Regularized Generative Adversarial Networks
• Chinese Discourse Segmentation Using Bilingual Discourse Commonality
• GANs beyond divergence minimization
• On Relativistic $f$-Divergences
• Trumpets: Injective Flows for Inference and Inverse Problems
• Metric Learning-based Generative Adversarial Network
• On Noise Injection in Generative Adversarial Networks
• Parallel and Limited Data Voice Conversion Using Stochastic Variational Deep Kernel Learning
• Artist Style Transfer Via Quadratic Potential
• Generating Magnetic Resonance Spectroscopy Imaging Data of Brain Tumours from Linear, Non-Linear and Deep Learning Models
• SDE approximations of GANs training and its long-run behavior
• BasketballGAN: Generating Basketball Play Simulation Through Sketching
• Image Generation Via Minimizing Fréchet Distance in Discriminator Feature Space
• A Generative Model for Hallucinating Diverse Versions of Super Resolution Images
• Learning a Hierarchical Latent-Variable Model of 3D Shapes
• Efficient Super Resolution For Large-Scale Images Using Attentional GAN
• Adversarial Binary Coding for Efficient Person Re-identification
• Updating the generator in PPGN-h with gradients flowing through the encoder
• Mode Penalty Generative Adversarial Network with adapted Auto-encoder
• Classification of sparsely labeled spatio-temporal data through semi-supervised adversarial learning
• Data-driven Accelerogram Synthesis using Deep Generative Models
• Data Augmentation of Room Classifiers using Generative Adversarial Networks
• GANE: A Generative Adversarial Network Embedding
• Reward-Reinforced Reinforcement Learning for Multi-agent Systems
• Unsupervised Object Matching for Relational Data
• Restricting Greed in Training of Generative Adversarial Network
• Semi-supervised mp-MRI Data Synthesis with StitchLayer and Auxiliary Distance Maximization
• A Self-Training Method for Semi-Supervised GANs
• Learning to Transport with Neural Networks
• LRS-DAG: Low Resource Supervised Domain Adaptation with Generalization Across Domains
• Distributed Function Minimization in Apache Spark
• Stochastic Bayesian Neural Networks
• Local AdaGrad-Type Algorithm for Stochastic Convex-Concave Optimization
• Momentum Contrastive Autoencoder: Using Contrastive Learning for Latent Space Distribution Matching in WAE
• Generative Adversarial Learning via Kernel Density Discrimination
• Generic Semi-Supervised Adversarial Subject Translation for Sensor-Based Human Activity Recognition
• Dist-GAN: An Improved GAN using Distance Constraints
• Cross-modal Image Retrieval with Deep Mutual Information Maximization
• Improving Stability of LS-GANs for Audio and Speech Signals
• Selective Sampling and Mixture Models in Generative Adversarial Networks
• On the Performance of Generative Adversarial Network (GAN) Variants: A Clinical Data Study
• Diversity Regularized Adversarial Learning
• An Interpretable Generative Model for Handwritten Digit Image Synthesis