Alternating the Population and Control Neural Networks to Solve High-Dimensional Stochastic Mean-Field Games
arXiv:2002.10113 · doi:10.1073/pnas.2024713118
Abstract
We present APAC-Net, an alternating population and agent control neural network for solving stochastic mean field games (MFGs). Our algorithm is geared toward high-dimensional instances of MFGs that are beyond reach with existing solution methods. We achieve this in two steps. First, we take advantage of the underlying variational primal-dual structure that MFGs exhibit and phrase it as a convex-concave saddle point problem. Second, we parameterize the value and density functions by two neural networks, respectively. By phrasing the problem in this manner, solving the MFG can be interpreted as a special case of training a generative adversarial network (GAN). We show the potential of our method on up to 100-dimensional MFG problems.
References in corpus (16)
- Neural Ordinary Differential Equations
- Stable Architectures for Deep Neural Networks
- A Machine Learning Framework for Solving High-Dimensional Mean Field Game and Mean Field Control Problems
- Game theoretic modelling of infectious disease dynamics and intervention methods: a mini-review
- FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Models
- Multi-Agent Reinforcement Learning via Double Averaging Primal-Dual Optimization
- Learning Generative Models with Sinkhorn Divergences
- A Mean-Field Optimal Control Formulation of Deep Learning
- On the Convergence and Robustness of Training GANs with Regularized Optimal Transport
- Linear-Quadratic Mean-Field Reinforcement Learning: Convergence of Policy Gradient Methods
- Discretize-Optimize vs. Optimize-Discretize for Time-Series Regression and Continuous Normalizing Flows
- The variational structure and time-periodic solutions for mean-field games systems
- Actor-Critic Provably Finds Nash Equilibria of Linear-Quadratic Mean-Field Games
- Discriminator optimal transport
- A mean-field game approach to price formation in electricity markets
- Splitting methods for a class of non-potential mean field games