When: Thursday 26th April 2018, 13:00-14:00.
Included in meeting:
The fortnightly BRAG meeting will be held tomorrow 26/04/18 at 1pm in room Y801. This week we will have presentations by Leah and Insha.
Title: A Brief Introduction to (Zero-Variance) Control Variates
A common objective in Bayesian inference is to estimate the expectation of a function with respect to the posterior. If approximate samples from the posterior are available then Monte Carlo integration can be used, with the quality of the estimator depending on the number of samples obtained. Control variates are a technique to lower the variance of the estimator without having to obtain additional samples. In this talk, I will explain the basics of control variates before going into more detail about a specific kind of control variates known as zero-variance control variates (ZV-CV). ZV-CV are applicable in applications where the derivatives of the log likelihood and log prior with respect to the parameters are available or can be unbiasedly estimated.
Title: Big data clustering
Abstract: I will be talking about some clustering methods (kmeans clustering and Dirichlet process Gaussian mixture models) I have found useful for clustering satellite images.
Farzana and Trish