RNNs Incrementally Evolving on an Equilibrium Manifold: A Panacea for Vanishing and Exploding Gradients?
Anil KagZiming ZhangVenkatesh SaligramaPublished in: ICLR (2020)
Keyphrases
- recurrent neural networks
- game theory
- nash equilibrium
- low dimensional
- feedforward neural networks
- nash equilibria
- high dimensional
- fuzzy neural network
- tangent space
- manifold learning
- euclidean space
- market equilibrium
- riemannian manifolds
- neural network
- variational inequalities
- genetic algorithm
- nonlinear dimensionality reduction
- echo state networks
- data sets