Wide Bayesian neural networks have a simple weight posterior: theory and accelerated sampling

Open in new window