C4I CENTER SEMINAR SERIES
Ryan Turner of the C4I Center presents
“The Diaconis and Freedman Example”
Diaconis and Freedman in a 1982 paper demonstrated an example of how density estimation using Dirichlet processes can be inconsistent even in a simple example. The paper is widely cited as a pitfall of Dirichlet processes or even nonparametric Bayesian inference. The paper is highly technical and does not perform any computational experiments. I will show some results of simulating the example in the paper, which results in some additional insights that resolves many issues in retrospect.
Ryan Turner is a recent Cambridge (UK) PhD, and now works on Bayesian nonparametric tracking at Northrop Grumman. His other publications are on his (outdated) Cambridge website, http://mlg.eng.cam.ac.uk/rdturner/publications.html.
1:00 pm - 2:00 pm
C4I Center ENGR 4705