Unrolled Generative Adversarial Networks

Abstract

We introduce a method to stabilize Generative Adversarial Networks (GANs) by defining the generator objective with respect to an unrolled optimization of the discriminator. This allows training to be adjusted between using the optimal discriminator in the generator's objective, which is ideal but infeasible in practice, and using the current value of the discriminator, which is often unstable and leads to poor solutions. We show how this technique solves the common problem of mode collapse, stabilizes training of GANs with complex recurrent generators, and increases diversity and coverage of the data distribution by the generator.

Cited by in corpus (153)

• Generative Adversarial Networks: An Overview
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Improved Training of Wasserstein GANs
• Objective-Reinforced Generative Adversarial Networks (ORGAN) for Sequence Generation Models
• How Generative Adversarial Networks and Their Variants Work: An Overview
• Generating Multi-label Discrete Patient Records using Generative Adversarial Networks
• Improved Precision and Recall Metric for Assessing Generative Models
• BAGAN: Data Augmentation with Balancing GAN
• GAN Lab: Understanding Complex Deep Generative Models using Interactive Visual Experimentation
• Stabilizing Training of Generative Adversarial Networks through Regularization
• The Numerics of GANs
• A Variational Inequality Perspective on Generative Adversarial Networks
• A Survey on Device Behavior Fingerprinting: Data Sources, Techniques, Application Scenarios, and Datasets
• The Mechanics of n-Player Differentiable Games
• Adversarial Feature Matching for Text Generation
• Deep generative models in inversion: a review and development of a new approach based on a variational autoencoder
• Data-driven modelling of nonlinear spatio-temporal fluid flows using a deep convolutional generative adversarial network
• Multi-Adversarial Domain Adaptation
• Photographic Image Synthesis with Cascaded Refinement Networks
• Triangle Generative Adversarial Networks
• Stabilizing Generative Adversarial Networks: A Survey
• DASVDD: Deep Autoencoding Support Vector Data Descriptor for Anomaly Detection
• Dual Adversarial Auto-Encoders for Clustering
• A Game Theoretic Framework for Model Based Reinforcement Learning
• LOGAN: Latent Optimisation for Generative Adversarial Networks
• Is Generator Conditioning Causally Related to GAN Performance?
• Dropout-GAN: Learning from a Dynamic Ensemble of Discriminators
• Distribution Matching in Variational Inference
• Prescribed Generative Adversarial Networks
• Improving GANs Using Optimal Transport
• Understanding Short-Horizon Bias in Stochastic Meta-Optimization
• Discrete Flows: Invertible Generative Models of Discrete Data
• Convergence of Learning Dynamics in Stackelberg Games
• Painting halos from cosmic density fields of dark matter with physically motivated neural networks
• Mode Seeking Generative Adversarial Networks for Diverse Image Synthesis
• Adversarial Training in Affective Computing and Sentiment Analysis: Recent Advances and Perspectives
• Semi-Amortized Variational Autoencoders
• Face Synthesis from Visual Attributes via Sketch using Conditional VAEs and GANs
• Generative Adversarial Networks (GANs): An Overview of Theoretical Model, Evaluation Metrics, and Recent Developments
• DiverGAN: An Efficient and Effective Single-Stage Framework for Diverse Text-to-Image Generation
• Adversarial score matching and improved sampling for image generation
• FedGAN: Federated Generative Adversarial Networks for Distributed Data
• Toward Diverse Text Generation with Inverse Reinforcement Learning
• End-to-End Learning for Structured Prediction Energy Networks
• Data-Driven Crowd Simulation with Generative Adversarial Networks
• Learning Generative Models across Incomparable Spaces
• FCC-GAN: A Fully Connected and Convolutional Net Architecture for GANs
• MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks
• Learning Texture Manifolds with the Periodic Spatial GAN
• Copulas as High-Dimensional Generative Models: Vine Copula Autoencoders
• Probabilistic Dual Network Architecture Search on Graphs
• Generative adversarial networks and adversarial methods in biomedical image analysis
• Learning Optimal Representations with the Decodable Information Bottleneck
• Age-Oriented Face Synthesis with Conditional Discriminator Pool and Adversarial Triplet Loss
• Some Theoretical Insights into Wasserstein GANs
• Imperfect ImaGANation: Implications of GANs Exacerbating Biases on Facial Data Augmentation and Snapchat Selfie Lenses
• CorGAN: Correlation-Capturing Convolutional Generative Adversarial Networks for Generating Synthetic Healthcare Records
• Interactive Sketch & Fill: Multiclass Sketch-to-Image Translation
• Negative Momentum for Improved Game Dynamics
• Generative Adversarial Networks (GANs): What it can generate and What it cannot?
• Learning Normal Dynamics in Videos with Meta Prototype Network
• Stein Neural Sampler
• Self-Diagnosing GAN: Diagnosing Underrepresented Samples in Generative Adversarial Networks
• AlignGAN: Learning to Align Cross-Domain Images with Conditional Generative Adversarial Networks
• Adversarial symmetric GANs: bridging adversarial samples and adversarial networks
• Multi-task Maximum Entropy Inverse Reinforcement Learning
• A New Semi-supervised Learning Benchmark for Classifying View and Diagnosing Aortic Stenosis from Echocardiograms
• Prediction Under Uncertainty with Error-Encoding Networks
• Cross-language Learning with Adversarial Neural Networks: Application to Community Question Answering
• Distributional Adversarial Networks
• MichiGAN: Multi-Input-Conditioned Hair Image Generation for Portrait Editing
• Composite Functional Gradient Learning of Generative Adversarial Models
• DA-GAN: Instance-level Image Translation by Deep Attention Generative Adversarial Networks (with Supplementary Materials)
• Relational Data Synthesis using Generative Adversarial Networks: A Design Space Exploration
• Newton-type Methods for Minimax Optimization
• Jointly Measuring Diversity and Quality in Text Generation Models
• GraN-GAN: Piecewise Gradient Normalization for Generative Adversarial Networks
• Alleviating Mode Collapse in GAN via Diversity Penalty Module
• TripletGAN: Training Generative Model with Triplet Loss
• Inception Score, Label Smoothing, Gradient Vanishing and -log(D(x)) Alternative
• CounteRGAN: Generating Realistic Counterfactuals with Residual Generative Adversarial Nets
• GDPP: Learning Diverse Generations Using Determinantal Point Process
• Interaction Matters: A Note on Non-asymptotic Local Convergence of Generative Adversarial Networks
• Predict and Constrain: Modeling Cardinality in Deep Structured Prediction
• Understanding the Effectiveness of Lipschitz-Continuity in Generative Adversarial Nets
• Understanding and Stabilizing GANs' Training Dynamics with Control Theory
• Gradient Descent-Ascent Provably Converges to Strict Local Minmax Equilibria with a Finite Timescale Separation
• Generative Minimization Networks: Training GANs Without Competition
• The Differentiable Cross-Entropy Method
• Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection
• Understanding Human Behaviors in Crowds by Imitating the Decision-Making Process
• CDE-GAN: Cooperative Dual Evolution Based Generative Adversarial Network
• xAI-GAN: Enhancing Generative Adversarial Networks via Explainable AI Systems
• Poincaré Recurrence, Cycles and Spurious Equilibria in Gradient-Descent-Ascent for Non-Convex Non-Concave Zero-Sum Games
• First Order Generative Adversarial Networks
• Improving Consistency and Correctness of Sequence Inpainting using Semantically Guided Generative Adversarial Network
• Convergence of Multi-Agent Learning with a Finite Step Size in General-Sum Games
• Measuring Fairness in Generative Models
• Snore-GANs: Improving Automatic Snore Sound Classification with Synthesized Data
• Learning Implicit Generative Models by Teaching Explicit Ones
• Taming GANs with Lookahead-Minmax
• Adaptive Weighted Discriminator for Training Generative Adversarial Networks
• NOVAS: Non-convex Optimization via Adaptive Stochastic Search for End-to-End Learning and Control
• GANs with Variational Entropy Regularizers: Applications in Mitigating the Mode-Collapse Issue
• Improved BiGAN training with marginal likelihood equalization
• Inclusive GAN: Improving Data and Minority Coverage in Generative Models
• Semi-Implicit Generative Model
• Normal Learning in Videos with Attention Prototype Network
• Sample weighting as an explanation for mode collapse in generative adversarial networks
• A Novel Framework for Selection of GANs for an Application
• DO-GAN: A Double Oracle Framework for Generative Adversarial Networks
• Diverse Image Generation via Self-Conditioned GANs
• MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining
• Convergence and Sample Complexity of SGD in GANs
• Generative Adversarial Networks and Adversarial Autoencoders: Tutorial and Survey
• On the Exploitation of Neuroevolutionary Information: Analyzing the Past for a More Efficient Future
• Adaptive Divergence for Rapid Adversarial Optimization
• Proximal Mapping for Deep Regularization
• CollaborER: A Self-supervised Entity Resolution Framework Using Multi-features Collaboration
• Benefiting Deep Latent Variable Models via Learning the Prior and Removing Latent Regularization
• Generative networks as inverse problems with fractional wavelet scattering networks
• Improving Text to Image Generation using Mode-seeking Function
• Stingray Detection of Aerial Images Using Augmented Training Images Generated by A Conditional Generative Model
• Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models
• SyMetric: Measuring the Quality of Learnt Hamiltonian Dynamics Inferred from Vision
• Out-of-Sample Testing for GANs
• Domain Partitioning Network
• IID-GAN: an IID Sampling Perspective for Regularizing Mode Collapse
• Minimax Optimization with Smooth Algorithmic Adversaries
• Encoder-Powered Generative Adversarial Networks
• Adversarial sampling of unknown and high-dimensional conditional distributions
• Teaching a GAN What Not to Learn
• Stochastic Deconvolutional Neural Network Ensemble Training on Generative Pseudo-Adversarial Networks
• On the potential for open-endedness in neural networks
• Adversarial Regularization as Stackelberg Game: An Unrolled Optimization Approach
• Coulomb Autoencoders
• Virtual Conditional Generative Adversarial Networks
• Entropy optimized semi-supervised decomposed vector-quantized variational autoencoder model based on transfer learning for multiclass text classification and generation
• Stability of Gradient Learning Dynamics in Continuous Games: Vector Action Spaces
• Normalized Diversification
• SVMax: A Feature Embedding Regularizer
• Style-Restricted GAN: Multi-Modal Translation with Style Restriction Using Generative Adversarial Networks
• SetGAN: Improving the stability and diversity of generative models through a permutation invariant architecture
• Processsing Simple Geometric Attributes with Autoencoders
• High Diversity Attribute Guided Face Generation with GANs
• CBP: Backpropagation with constraint on weight precision using a pseudo-Lagrange multiplier method
• Reparameterized Sampling for Generative Adversarial Networks
• Instance-Conditioned GAN
• When Relation Networks meet GANs: Relation GANs with Triplet Loss
• Dist-GAN: An Improved GAN using Distance Constraints
• Human Annotations Improve GAN Performances
• Semi-Supervised Learning with Meta-Gradient
• IllumiNet: Transferring Illumination from Planar Surfaces to Virtual Objects in Augmented Reality