Dynamic Game Theoretic Neural Optimizer
arXiv:2105.03788
Abstract
The connection between training deep neural networks (DNNs) and optimal control theory (OCT) has attracted considerable attention as a principled tool of algorithmic design. Despite few attempts being made, they have been limited to architectures where the layer propagation resembles a Markovian dynamical system. This casts doubts on their flexibility to modern networks that heavily rely on non-Markovian dependencies between layers (e.g. skip connections in residual networks). In this work, we propose a novel dynamic game perspective by viewing each layer as a player in a dynamic game characterized by the DNN itself. Through this lens, different classes of optimizers can be seen as matching different types of Nash equilibria, depending on the implicit information structure of each (p)layer. The resulting method, called Dynamic Game Theoretic Neural Optimizer (DGNOpt), not only generalizes OCT-inspired optimizers to richer network class; it also motivates a new training principle by solving a multi-player cooperative game. DGNOpt shows convergence improvements over existing methods on image classification datasets with residual and inception networks. Our work marries strengths from both OCT and game theory, paving ways to new algorithmic opportunities from robust optimal control and bandit-based optimization.
Accepted in International Conference on Machine Learning (ICML) 2021 as Oral
References in corpus (4)
- Deep Learning Theory Review: An Optimal Control and Dynamical Systems Perspective
- Analysing Neural Network Topologies: a Game Theoretic Approach
- Measurements of Three-Level Hierarchical Structure in the Outliers in the Spectrum of Deepnet Hessians
- Mean-field Langevin System, Optimal Control and Deep Neural Networks