Fast Black-box Variational Inference through Stochastic Trust-Region Optimization

Jeffrey Regier, Michael I. Jordan, Jon McAuliffe

Neural Information Processing Systems 

We introduce TrustVI, a fast second-order algorithm for black-box variational inference based on trust-region optimization and the "reparameterization trick." At each iteration, TrustVI proposes and assesses a step based on minibatches of draws from the variational distribution.