American Statistical Association
I will give a presentation on the Dirichlet Process. There were several unresolved issues in my previous seminar on the Gershman and Blei (2012) paper. They explained the remarkable Chinese Restaurant Process (CRP) construction method for the Dirichlet Process. The CRP allows a statistical model to accommodate an infinite number of parameters. Concerns over model overfit are greatly reduced. However, many details remain unexplained, such as what the DP actually looks like and why the CRP must loop through all N observations?
Since then my colleagues and I have been filling in the omitted details. We have worked out key results in the literature that are either hidden away or assumed understood. We also wrote a crude but practical program in R to help visualize the DP. They will be covered in my talk. Examples of DP mixture modeling will be given using patient-reported outcomes data. Our overall intent is to explain DP clearly so that you may feel encouraged to try Bayesian Nonparametric methods in your own research.
|Date:||Wednesday, March 21, 2018 - CANCELLED DUE TO WEATHER CONDITIONS|
|Time:||4:00 - 5:00 P.M.|
Memorial Sloan Kettering Cancer Center
Department of Epidemiology and Biostatistics
485 Lexington Avenue
(Between 46th & 47th Streets)
2nd Floor, Conference Room B
New York, New York
**Outside visitors please email email@example.com for building access. You must be on the security list to enter the floor.