20
PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009.

PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Embed Size (px)

Citation preview

Page 1: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PARTICLE LEARNINGA semester later

Hedibert Freitas Lopes

February 19th 2009.

Page 2: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Group meetings

Discussion of Storvik and Liu and West (LW) papersCreation of research sub-groupsKernel choice in LW scheme (Petris)APF, SIR & LW (Lopes)Nonlinear PL (Polson)LW + jittering SS (Fearnhead)SMC for long memory time series models (Macaro)SMC for DSGE models (Petralia)PL in structured AR models (Prado)Adaptive SMC in Mixture Analysis (Taylor)SMC for long memory time series models (Macaro)

Page 3: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Sequential Importance Sampling

Page 4: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Particle degeneracy

Page 5: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PL scheme

Page 6: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

No degeneracy

Page 7: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Resample-propagate or propagate-resample?

Page 8: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009
Page 9: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Sufficient statistics

Page 10: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PL versus LW

Page 11: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PL versus MCMC

Page 12: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Smoothing

Page 13: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 1: PL in structured AR models

Prado & Lopes (2009)

Page 14: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 2: SMC in LMSV models

Macaro & Lopes (2009)

Page 15: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 3: Combining PL and LW

Petralia, Hao, Carvalho and Lopes (2009)DGSE : Dynamic General Stochastic Equilibrium

Page 16: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 4: PL in DGSE models

Niemi, Chiranjit, Carvalho & Lopes (2009)

Page 17: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 5: PL in epidemic SEIR models

Dukic, Lopes & Polson (2009)SEIR: susceptible exposed infected recovered

Page 18: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

PROJECT 6: PL in dynamic factor models

Lopes (2009)

Page 19: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Joint Statistical Meetings 2009

Invited SessionHedibert Lopes – Particle Learning and Smoothing

Topic Contributed Session – “Particle Learning”Raquel Prado – PL for Autoregressive Models with Structured PriorsChiranjit Mukherjee – PL Without Conditional Sufficient StatisticsChristian Macaro – PL for Long Memory Stochastic Volatility Models

Contributed SessionFrancesca Petralia – PL for Dynamic Stochastic General Equilibrium Models

Page 20: PARTICLE LEARNING A semester later Hedibert Freitas Lopes February 19 th 2009

Other projects