Bayesian Inference in Single-Layer Neural Networks
Robert Paige, (Texas Tech University), firstname.lastname@example.org
Approximate marginal Bayesian computation and inference are developed for single-layer neural network models. In particular we consider the classical neural network which uses basis functions and a novel wavelet neural network with wavelet basis functions. The marginal considerations include determination of approximate Bayes factors for model choice, approximate predictive density computation for a future observable, and determination of approximate Bayes estimates for the nonlinear regression function. Standard conjugate analysis applied to the linear parameters leads to an explicit posterior on the nonlinear parameters. The proposed methodology is illustrated in the context of a nonlinear dataset which involves a univariate nonlinear regression model.