|
Activity Number:
|
440
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #301595 |
|
Title:
|
On Real-parameter Evolutionary Monte Carlo Algorithm
|
|
Author(s):
|
Gopika R. Goswami*+
|
|
Companies:
|
Harvard University
|
|
Address:
|
Science Center, 7th Floor, Cambridge, MA, 02138,
|
|
Keywords:
|
exchange operation ; cross-over operation ; Markov chain Monte Carlo ; Real-parameter Evolutionary Monte Carlo ; snooker algorithm ; temperature ladder
|
|
Abstract:
|
Real-parameter Evolutionary Monte Carlo algorithm (EMC) has been proposed as an effective method not only for sampling from multi-modal distributions but also for stochastic optimization. The authors have shown by various well-chosen examples that it works better than parallel tempering, it has a certain ability to learn from the past and it improves mixing by sampling along a temperature ladder. Our goal is to improve the performance of this method and provide some practical implementation guidelines. First, we introduce some new exchange moves which equips the algorithm with more learning capability and makes it faster while preserving the quality of the samples (as measured by integrated autocorrelation time) produced from the target distribution. Second, we introduce some premiliminary run based tools for contructing the temperature ladder. Lastly, we generalize a result originally proved about the validity of the conditional (or line) sampling step in the context of the snooker algorithm which is a type of move involved in the EMC.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2004 program |
