Echo State Networks trained by Tikhonov least squares are L2(μ) approximators of ergodic dynamical systems.
Allen G. HartJames L. HookJonathan H. P. DawesPublished in: CoRR (2020)
Keyphrases
- dynamical systems
- echo state networks
- least squares
- recurrent neural networks
- differential equations
- tikhonov regularization
- dynamic systems
- nonlinear dynamical systems
- markov chain
- control theory
- linear systems
- state space
- qualitative simulation
- stationary distribution
- optical flow
- dynamical behavior
- regularization parameter
- predictive state representations
- phase space
- image restoration
- neural network
- discrete dynamical systems
- ls svm
- multi agent
- reinforcement learning