Please use this identifier to cite or link to this item:
https://scholarbank.nus.edu.sg/handle/10635/122666
Title: | NON-STATIONARY GAUSSIAN PROCESS MODELS FOR DETERMINISTIC AND STOCHASTIC SIMULATION | Authors: | YANG LINCHANG | Keywords: | Metamodels, Gaussian process, Non-stationary, Space mapping, Heteroscedastic, Variational approximation | Issue Date: | 31-Jul-2015 | Citation: | YANG LINCHANG (2015-07-31). NON-STATIONARY GAUSSIAN PROCESS MODELS FOR DETERMINISTIC AND STOCHASTIC SIMULATION. ScholarBank@NUS Repository. | Abstract: | TO HELP THE ANALYSIS OF EXPENSIVE SIMULATIONS, RESEARCHERS START TO USE METAMODELS TO APPROXIMATE THE OUTPUT BASED ON FINITE SIMULATED RESULTS. ONE POPULAR METAMODEL IS GAUSSIAN PROCESS BUT IT ASSUMES THAT THE COVARIANCE IS STATIONARY SO IT PERFORMS POORLY IN DEALING WITH NON-STATIONARY DATA. IN THIS THESIS, WE PROPOSE TWO NON-STATIONARY GAUSSIAN PROCESS MODELS. ONE IS FOR THE DETERMINISTIC CASE AND THE OTHER IS FOR THE STOCHASTIC CASE. THE DETERMINISTIC MODEL CONVERTS A NON-STATIONARY PROBLEM INTO A STATIONARY ONE VIA SPACE MAPPING. IT COULD PROVIDE CONTINUOUS RESULTS AND REASONABLE MAPPING RULES. THE STOCHASTIC MODEL USES THE VARIATIONAL APPROXIMATION TO ESTIMATE THE PARAMETER AND IS PRACTICAL EVEN WITHOUT REPLICATION. MOREOVER, SEQUENTIAL DESIGN METHODS FOR BOTH MODELS ARE GIVEN TO ALLOCATE THE SIMULATION BUDGET WITH PROPOSED MODELS AND INTEGRATION STRATEGIES OF COMBINING RESULTS OF DIFFERENT SIMULATION FIDELITIES ARE INTRODUCED SEPARATELY FOR EACH MODEL. | URI: | http://scholarbank.nus.edu.sg/handle/10635/122666 |
Appears in Collections: | Ph.D Theses (Open) |
Show full item record
Files in This Item:
File | Description | Size | Format | Access Settings | Version | |
---|---|---|---|---|---|---|
YANGLC.pdf | 2.89 MB | Adobe PDF | OPEN | None | View/Download |
Google ScholarTM
Check
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.