Global Convergence to the Equilibrium of GANs using Variational Inequalities
arXiv:1808.01531
Abstract
In optimization, the negative gradient of a function denotes the direction of steepest descent. Furthermore, traveling in any direction orthogonal to the gradient maintains the value of the function. In this work, we show that these orthogonal directions that are ignored by gradient descent can be critical in equilibrium problems. Equilibrium problems have drawn heightened attention in machine learning due to the emergence of the Generative Adversarial Network (GAN). We use the framework of Variational Inequalities to analyze popular training algorithms for a fundamental GAN variant: the Wasserstein Linear-Quadratic GAN. We show that the steepest descent direction causes divergence from the equilibrium, and convergence to the equilibrium is achieved through following a particular orthogonal direction. We call this successful technique Crossing-the-Curl, named for its mathematical derivation as well as its intuition: identify the game's axis of rotation and move "across" space in the direction towards smaller "curling".
References in corpus (14)
- Progressive Growing of GANs for Improved Quality, Stability, and Variation
- Energy-based Generative Adversarial Network
- Which Training Methods for GANs do actually Converge?
- f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
- Generative Adversarial Imitation Learning
- The Numerics of GANs
- A Variational Inequality Perspective on Generative Adversarial Networks
- The Mechanics of n-Player Differentiable Games
- NP-hardness of Deciding Convexity of Quartic Polynomials and Related Problems
- Generative Adversarial Nets from a Density Ratio Estimation Perspective
- Understanding GANs: the LQG Setting
- McGan: Mean and Covariance Feature Matching GAN
- Online Monotone Games
- Online Monotone Optimization
Cited by in corpus (17)
- LOGAN: Latent Optimisation for Generative Adversarial Networks
- Competitive Gradient Descent
- Stochastic Variance Reduction for Variational Inequality Methods
- On Solving Minimax Optimization Locally: A Follow-the-Ridge Approach
- Finding Mixed Nash Equilibria of Generative Adversarial Networks
- Implicit competitive regularization in GANs
- Training Generative Adversarial Networks by Solving Ordinary Differential Equations
- Stable Opponent Shaping in Differentiable Games
- Smooth markets: A basic mechanism for organizing gradient-based learners
- Latent-Optimized Adversarial Neural Transfer for Sarcasm Detection
- Convergence and Sample Complexity of SGD in GANs
- Competitive Mirror Descent
- Competitive Policy Optimization
- Causal Inference in Network Economics
- Bridging Explicit and Implicit Deep Generative Models via Neural Stein Estimators
- Exploiting the Hidden Tasks of GANs: Making Implicit Subproblems Explicit
- LRS-DAG: Low Resource Supervised Domain Adaptation with Generalization Across Domains