<?xml version='1.0' encoding='UTF-8'?><?xml-stylesheet href='static/style.xsl' type='text/xsl'?><OAI-PMH xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/ http://www.openarchives.org/OAI/2.0/OAI-PMH.xsd"><responseDate>2026-11-22T08:53:14Z</responseDate><request verb="GetRecord" identifier="oai:ecommons.cornell.edu:1813/11103" metadataPrefix="dim">https://ecommons.cornell.edu/server/oai/request</request><GetRecord><record><header><identifier>oai:ecommons.cornell.edu:1813/11103</identifier><datestamp>2026-05-13T16:36:38Z</datestamp><setSpec>com_1813_35</setSpec><setSpec>col_1813_47</setSpec></header><metadata><dim:dim xmlns:dim="http://www.dspace.org/xmlns/dspace/dim" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:doc="http://www.lyncode.com/xoai" xsi:schemaLocation="http://www.dspace.org/xmlns/dspace/dim http://www.dspace.org/schema/dim.xsd">
   <dim:field mdschema="dc" element="contributor" qualifier="author">Bliznyuk, Nikolay</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="accessioned">2008-07-10T15:44:07Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="available">2013-07-10T06:21:14Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="issued">2008-07-10T15:44:07Z</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="uri">https://hdl.handle.net/1813/11103</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="bibid">6397183</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="abstract" lang="en_US">Markov Chain Monte Carlo (MCMC) is nowadays a standard&#xd;
approach to numerical computation of integrals of the&#xd;
posterior density $\pi$ of the parameter vector $\eta$.&#xd;
Unfortunately, Bayesian inference using MCMC is&#xd;
computationally intractable when $\pi$ is expensive to&#xd;
evaluate. In this work, we develop practical methods that&#xd;
approximate $\pi$ with radial basis functions (RBFs) and&#xd;
Gaussian processes (GPs) interpolants and use the resulting&#xd;
cheap-to-evaluate surfaces in MCMC.  &#xd;
&#xd;
In Chapter 1, $\pi$&#xd;
arises from a nonlinear regression model with transformation&#xd;
and dependence. To build the RBF approximation, we limit&#xd;
evaluation of the computationally expensive regression&#xd;
function to points chosen on a high posterior density (HPD)&#xd;
region found using a local quadratic approximation of&#xd;
$\log(\pi)$ at its mode. We illustrate our approach on&#xd;
simulated data for a pollutant diffusion problem and study&#xd;
the frequentist coverage properties of credible intervals.&#xd;
&#xd;
In Chapter 2, we relax the assumptions about $\pi$ made in&#xd;
Chapter 1 and develop a derivative-free procedure GRIMA&#xd;
to approximate the logarithm of $\pi$ using RBF&#xd;
interpolation  over a HPD region of $\pi$ estimated using&#xd;
the RBF surface. We use GRIMA for Bayesian inference in a&#xd;
computationally intensive nonlinear regression model for&#xd;
real measured streamflow data in the Town Brook watershed.&#xd;
&#xd;
In Chapter 3, we study statistical models where it is&#xd;
possible to identify a minimal subvector $\beta$ of $\eta$&#xd;
responsible for the expensive computation in the evaluation&#xd;
of $\pi$. We propose two approaches to approximate $\pi$ by&#xd;
interpolation that exploit this computational structure. Our&#xd;
primary contribution is derivation of a GP interpolant that&#xd;
provably improves over some of the existing approaches by&#xd;
reducing the effective dimension of the interpolation&#xd;
problem from $\dim(\eta)$ to $\dim(\beta)$. When $\dim(\eta)&#xd;
$ is high but $\dim(\beta)$ is low, this allows one to&#xd;
dramatically reduce the number of expensive evaluations&#xd;
necessary to construct an accurate approximation of $\pi$.&#xd;
&#xd;
Our experiments indicate that our methods produce results similar to those&#xd;
when the true expensive posterior density is sampled by MCMC while reducing&#xd;
computational costs by well over an order of magnitude.</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="sponsorship" lang="en_US">NSF DMS-04-538</dim:field>
   <dim:field mdschema="dc" element="language" qualifier="iso" lang="en_US">en_US</dim:field>
   <dim:field mdschema="dc" element="relation" qualifier="isformatof">bibid: 6397183</dim:field>
   <dim:field mdschema="dc" element="subject" lang="en_US">inverse problems</dim:field>
   <dim:field mdschema="dc" element="subject" lang="en_US">response surface</dim:field>
   <dim:field mdschema="dc" element="title" lang="en_US">Posterior Approximation by Interpolation for Bayesian Inference in Computationally Expensive Statistical Models</dim:field>
   <dim:field mdschema="dc" element="type" lang="en_US">dissertation or thesis</dim:field>
   <dim:field mdschema="dc" element="format" qualifier="mimetype">application/pdf</dim:field>
   <dim:field mdschema="dspace" element="entity" qualifier="type">Publication</dim:field>
   <dim:field mdschema="cris" element="virtual" qualifier="collection" authority="https://cornell-ecommons.eks.prod.4science.cloud/handle/1813/47" confidence="600">Cornell Theses and Dissertations</dim:field>
   <dim:field mdschema="cris" element="virtual" qualifier="author">Bliznyuk, Nikolay</dim:field>
   <dim:field mdschema="cris" element="virtualsource" qualifier="collection">5893a6ea-7af3-41d7-abc6-04bcd26ab5df</dim:field>
   <dim:field mdschema="others" element="access-status">open.access</dim:field>
   <dim:field mdschema="others" element="access-status">open.access</dim:field>
   <dim:field mdschema="cerif" element="openaire" authority="" confidence="-1">&lt;Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="24398f22-934a-4d51-97ee-32c369adabce">
	&lt;Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843&lt;/Type>
	&lt;Language>en_US&lt;/Language>
   	&lt;Title>Posterior Approximation by Interpolation for Bayesian Inference in Computationally Expensive Statistical Models&lt;/Title>
   	&lt;PublishedIn>
    	&lt;Publication>
      	&lt;/Publication>
   	&lt;/PublishedIn>
   	&lt;PublicationDate>2008-07-10T15:44:07Z&lt;/PublicationDate>
   	&lt;Authors>
      	&lt;Author>
        	&lt;DisplayName>Bliznyuk, Nikolay&lt;/DisplayName>
         	&lt;Affiliation>
         		&lt;OrgUnit>
         		&lt;/OrgUnit>
         	&lt;/Affiliation>
      	&lt;/Author>
	&lt;/Authors>
   	&lt;Editors>
	&lt;/Editors>
    &lt;Publishers>
        &lt;Publisher>
            &lt;OrgUnit />
        &lt;/Publisher>
    &lt;/Publishers>
    &lt;Keyword>inverse problems&lt;/Keyword>
    &lt;Keyword>response surface&lt;/Keyword>
   	&lt;Abstract>Markov Chain Monte Carlo (MCMC) is nowadays a standard&#xd;
approach to numerical computation of integrals of the&#xd;
posterior density $\pi$ of the parameter vector $\eta$.&#xd;
Unfortunately, Bayesian inference using MCMC is&#xd;
computationally intractable when $\pi$ is expensive to&#xd;
evaluate. In this work, we develop practical methods that&#xd;
approximate $\pi$ with radial basis functions (RBFs) and&#xd;
Gaussian processes (GPs) interpolants and use the resulting&#xd;
cheap-to-evaluate surfaces in MCMC.  &#xd;
&#xd;
In Chapter 1, $\pi$&#xd;
arises from a nonlinear regression model with transformation&#xd;
and dependence. To build the RBF approximation, we limit&#xd;
evaluation of the computationally expensive regression&#xd;
function to points chosen on a high posterior density (HPD)&#xd;
region found using a local quadratic approximation of&#xd;
$\log(\pi)$ at its mode. We illustrate our approach on&#xd;
simulated data for a pollutant diffusion problem and study&#xd;
the frequentist coverage properties of credible intervals.&#xd;
&#xd;
In Chapter 2, we relax the assumptions about $\pi$ made in&#xd;
Chapter 1 and develop a derivative-free procedure GRIMA&#xd;
to approximate the logarithm of $\pi$ using RBF&#xd;
interpolation  over a HPD region of $\pi$ estimated using&#xd;
the RBF surface. We use GRIMA for Bayesian inference in a&#xd;
computationally intensive nonlinear regression model for&#xd;
real measured streamflow data in the Town Brook watershed.&#xd;
&#xd;
In Chapter 3, we study statistical models where it is&#xd;
possible to identify a minimal subvector $\beta$ of $\eta$&#xd;
responsible for the expensive computation in the evaluation&#xd;
of $\pi$. We propose two approaches to approximate $\pi$ by&#xd;
interpolation that exploit this computational structure. Our&#xd;
primary contribution is derivation of a GP interpolant that&#xd;
provably improves over some of the existing approaches by&#xd;
reducing the effective dimension of the interpolation&#xd;
problem from $\dim(\eta)$ to $\dim(\beta)$. When $\dim(\eta)&#xd;
$ is high but $\dim(\beta)$ is low, this allows one to&#xd;
dramatically reduce the number of expensive evaluations&#xd;
necessary to construct an accurate approximation of $\pi$.&#xd;
&#xd;
Our experiments indicate that our methods produce results similar to those&#xd;
when the true expensive posterior density is sampled by MCMC while reducing&#xd;
computational costs by well over an order of magnitude.&lt;/Abstract>
	&lt;Access xmlns="http://purl.org/coar/access_right" 
    >
    &lt;/Access>
&lt;/Publication>
</dim:field>
</dim:dim>
</metadata></record></GetRecord></OAI-PMH>