Stat 376: Spring 2008

Stat 376 Lecture schedule

Week 1: Review and basics: Numerical integration, MCMC and convergence assessment
Weeks 2-3: Theory: convergence and efficiency
Week 4: Semi- and non-parameteric models: Mixtures, neural network and kernel regressions, Dirchlet processes
Weeks 5-7 Advanced MCMC: Auxiliary variable techniques, parameter expansion, generalized ensembles
Week 8-9: Theory: convergence rates and bounds, algorithmic complexity
Week 10: Adaptive MC and stochastic approximation
Week 11: Marginal likelihoods and normalizing constants
Week 12-14: Variational methods, graph propagation
Somewhere: Sequential MC; optimization techniques

Week	Date	Lecture Topics	Reading (* required)
1	1/15 1/17	Course overview; review of quadrature, MC integration Review of finite Markov chain theory	Background reading in Ripley, Robert & Casella Review papers
2	1/22 1/24	Notions of efficiency in MC integration and MCMC Convergence diagnostics (discussion)	Cowles & Carlin (1996)^* Convergence diagnostics
3	1/29 1/31	Gibbs fields; Ising, Potts, spinglass models; Neural networks Tempering and annealing	Tutorial notes^* Tempering and annealing
4	2/5 2/7	Monte Carlo optimization and MC likelihood Parameter estimation in Gibbs measures	Geyer & Thompson (1992)^, Geyer (1991) Besag (1975)^, Moller (2006), Others
5	2/12 2/14	Estimating marginal likelihoods/normalizing constants Reversible jump and transdimensional MCMC	Gelman & Meng (1998), Sinharay & Stern (2005), Raftery & Newton (2006) More on marginal likelihoods Sisson (2005)^*, Green (1995) More
6	2/19 2/21	Convergence rates and statistical efficiency Statistical efficiency and algorithmic complexity	Jones & Hobert (2001)^* Roberts & Rosenthal (2004) Lecture notes
7	2/26 2/28	Markov chains in general state spaces Langevin diffusion and SDEs, hybrid MC	Roberts & Rosenthal (1997)^* Tierney (1994) (w/ disc) Roberts & Rosenthal (2001)^* Langevin algorithms
8	3/4 3/6	Dirichlet processes and mixture models	Muller & Quintana (2004)^* Dirichlet processes