PAC-Bayesian Error Bound, via Rényi Divergence, for a Class of Linear Time-Invariant State-Space Models

11 Mihály Petreczky PAC-Bayesian Error Bound, via Rényi Divergence, for a Class 2 let h be such that LN (h ) + regularization term is small. Question: What can we say about the true error L(h ) ? /11 Mihály Petreczky PAC-Bayesian Error Bound, via Rényi Divergence, for a Class Recurrent neural networks (RNNs) with a linear activation function, and classical autoregressive models (ARX, ARMAX) are included. /11 Mihály Petreczky PAC-Bayesian Error Bound, via Rényi Divergence, for a Class The data is generated by (2) with 2 states, such that nu = ny = 1, eg (t) ∼ N (0, Qe ), hypotheses: linear systems with two states.