Activity Number:
|
23
|
Type:
|
Topic Contributed
|
Date/Time:
|
Sunday, July 31, 2016 : 2:00 PM to 3:50 PM
|
Sponsor:
|
International Society for Bayesian Analysis (ISBA)
|
Abstract #320743
|
|
Title:
|
Efficient Importance Sampling Methods for Estimating Parameters in SGLMMs and Improving Prediction
|
Author(s):
|
Vivekananda Roy*
|
Companies:
|
Iowa State University
|
Keywords:
|
Importance sampling ;
Markov chain Monte Carlo ;
Relevance vector machine ;
Support vector machine
|
Abstract:
|
Importance sampling is a Monte Carlo method where samples from one distribution are used to estimate expectations with respect to others. The naive importance sampling estimator with a single importance density can be unstable. We consider effective multiple importance sampling (MIS) estimators where samples from multiple distributions are combined for efficiently estimating means and (ratios of) normalizing constants of a large number of probability distributions. In massive dimensional complex data sets, reproducing kernel Hilbert space based regression methods can be used to find an appropriate function of the covariates to predict the response variables. We discuss the use of MIS methods for improved inference and prediction utilizing relevance vector machine models. Our methods are illustrated with simulation and real data examples.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2016 program
|