Stochastic Variational Deep Kernel Learning - NIPS 2016

Stochastic Variational Deep Kernel Learning NIPS 2016 Paper: Code: Authors: Andrew Gordon Wilson*, Zhiting Hu*, Ruslan Salakhutdinov, Eric P. Xing This work can be used as a plug-in to stand-alone deep networks, with minor additional runtime overhead, in exchange for improved predictive performance, interpretability, and full predictive distributions. SV-DKL exploits algebraic structure in deep kernels formed from (e.g. convoluti