Colin_slides DP

Colin_slides DP - Markov Chain Sampling Methods for...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon
Markov Chain Sampling Methods for Dirichlet Process Mixture Models Radford M. Neal, University of Toronto, Ontario, Canada Presented by Colin DeLong
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Outline z Introduction z Dirichlet process mixture models z Gibbs sampling w/ conjugate priors z Algorithms 1, 2, and 3 z Methods for handling non-conjugate priors z Algorithm 4 z Metropolis-Hastings and partial Gibbs z Algorithms 5, 6, and 7 z Gibbs sampling w/ auxiliary parameters z Algorithm 8 z Experiments (well, one)
Background image of page 2
Introduction z Some problems are more accurately represented with non-conjugate priors z z Climatology opinion quantification (Al-Awadhi & Garthwaite, 2001) z z Non-conjugate priors + Gibbs = headache. z Update integrals are nasty to compute z Solution? Metropolis-Hastings + partial Gibbs.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
models z Basic idea z Given data y 1 ,…, y n ind. drawn from an unknown distribution ( y i may be multivariate) z Model the unknown distribution as being drawn from of a mixture of distributions F ( θ ), w/ mixing distribution over θ being
Background image of page 4
Image of page 5
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 21

Colin_slides DP - Markov Chain Sampling Methods for...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online