Quasi-Newton Quasi-Monte Carlo for variational Bayes

Sifan Liu, Art B. Owen

Many machine learning problems optimize an objective that must be measured with noise. The primary method is a first order stochastic gradient descent using one or more Monte Carlo…