Quick and dirty torch implementation of the network featured in Scalable Bayesian optimization using deep neural networks by Snoek et al. (ICML 2015).
Note on hyperparameters: the alpha
and beta
hypers are neither fit or
inferred here because the marginal likelihood (Eq. 8) and its gradient still
needs to be implemented.