Complex Momentum for Optimization in Games

Abstract

We generalize gradient descent with momentum for optimization in differentiable games to have complex-valued momentum. We give theoretical motivation for our method by proving convergence on bilinear zero-sum games for simultaneous and alternating updates. Our method gives real-valued parameter updates, making it a drop-in replacement for standard optimizers. We empirically demonstrate that complex-valued momentum can improve convergence in realistic adversarial games - like generative adversarial networks - by showing we can find better solutions with an almost identical computational cost. We also show a practical generalization to a complex-valued Adam variant, which we use to train BigGAN to better inception scores on CIFAR-10.

References in corpus (42)

• Adam: A Method for Stochastic Optimization
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Which Training Methods for GANs do actually Converge?
• Meta-Learning for Semi-Supervised Few-Shot Classification
• Emergent Tool Use From Multi-Agent Autocurricula
• Lookahead Optimizer: k steps forward, 1 step back
• On Empirical Comparisons of Optimizers for Deep Learning
• The Numerics of GANs
• A Variational Inequality Perspective on Generative Adversarial Networks
• Solving a Class of Non-Convex Min-Max Games Using Iterative First Order Methods
• Connecting Generative Adversarial Networks and Actor-Critic Methods
• Stochastic Hyperparameter Optimization through Hypernetworks
• A Game Theoretic Framework for Model Based Reinforcement Learning
• Reducing Noise in GAN Training with Variance Reduced Extragradient
• LOGAN: Latent Optimisation for Generative Adversarial Networks
• Open-ended Learning in Symmetric Zero-sum Games
• Hamiltonian Descent Methods
• Differentiable Game Mechanics
• A Closer Look at the Optimization Landscapes of Generative Adversarial Networks
• Understanding Neural Architecture Search Techniques
• Self-Tuning Networks: Bilevel Optimization of Hyperparameters using Structured Best-Response Functions
• Linear Lower Bounds and Conditioning of Differentiable Games
• Multi-objective training of Generative Adversarial Networks with multiple discriminators
• Towards Better Understanding of Adaptive Gradient Algorithms in Generative Adversarial Nets
• Accelerating Smooth Games by Manipulating Spectral Shapes
• From Poincaré Recurrence to Convergence in Imperfect Information Games: Finding Equilibrium via Regularization
• Implicit competitive regularization in GANs
• Stochastic Hamiltonian Gradient Methods for Smooth Games
• Convergence of Gradient Methods on Bilinear Zero-Sum Games
• Training Generative Adversarial Networks by Solving Ordinary Differential Equations
• A Unified Analysis of First-Order Methods for Smooth Games via Integral Quadratic Constraints
• Adversarial Example Games
• Newton-type Methods for Minimax Optimization
• Real World Games Look Like Spinning Tops
• On the Suboptimality of Negative Momentum for Minimax Optimization
• Policy-Aware Model Learning for Policy Gradient Methods
• Optimality and Stability in Non-Convex Smooth Games
• Teaching with Commentaries
• Competitive Mirror Descent
• A Limited-Capacity Minimax Theorem for Non-Convex Games or: How I Learned to Stop Worrying about Mixed-Nash and Love Neural Nets
• Average-case Acceleration for Bilinear Games and Normal Matrices
• Self-Tuning Stochastic Optimization with Curvature-Aware Gradient Filtering