<?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>2026-11-22T08:51:27Z</responseDate><request verb="GetRecord" identifier="oai:ecommons.cornell.edu:1813/59449" metadataPrefix="dim">https://ecommons.cornell.edu/server/oai/request</request><GetRecord><record><header><identifier>oai:ecommons.cornell.edu:1813/59449</identifier><datestamp>2026-05-15T19:47:52Z</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">Xue, Yexiang</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="chair">Selman, Bart</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="committeeMember">Gomes, Carla P.</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="committeeMember">Joachims, Thorsten</dim:field>
   <dim:field mdschema="dc" element="contributor" qualifier="committeeMember">Hopcroft, John E.</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="accessioned">2018-10-23T13:23:08Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="available">2020-06-04T06:01:59Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="issued">2018-05-30</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="other">ProQuest Submission ID: 10840</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="other">ProQuest Publication ID: 10816540</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="uri">https://hdl.handle.net/1813/59449</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="doi">https://doi.org/10.7298/X4V986B4</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="bibid">10489534</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="abstract">Problems at the intersection of reasoning, optimization, and learning often involve multi-stage inference. For example, making decisions based on machine learning models often leads to multi-stage inference problems where probabilistic models learned from data are embedded as first-stage subproblems within a global second-stage problem for decision-making. Multi-agent reasoning also involves multi-stage inference, since the reasoning of any given agent has to incorporate the goals of the other agents. As a result, the decision processes of the other agents are embedded as first-stage subproblems within the overall decision-making problem of that agent. With formal complexities beyond NP, multi-stage inference problems are often highly intractable. In this thesis, I introduce a novel computational framework, based on embeddings, to tackle multi-stage inference problems. Our embedding technique approximates the intractable sub-problems of a multi-stage inference problem through a series of novel representations. We then embed these representations into the global problem, effectively reducing a multi-stage inference to a single-stage inference. As one example, I present a novel way to encode the reward allocation problem for a two-stage organizer--agent game as a single-stage optimization. The encoding embeds an approximation of the agents' decision-making process into the organizer's problem in the form of linear constraints. We apply this methodology to eBird, a well-established citizen-science program for collecting bird observations, in a game called Avicaching. Our AI-based reward allocation was shown to be highly effective, surpassing the expectations of the eBird organizers and bird conservation experts. As another example, I present a novel constant approximation algorithm to solve stochastic optimization problems which identify the optimal policy that maximizes the expectation of a stochastic objective. To tackle this problem, I propose the embedding of its intractable counting subproblems as queries to NP oracles subject to additional XOR constraints. As a result, the entire problem is encoded as a single NP-equivalent optimization. This approach outperforms state-of-the-art solvers based on variational inference and MCMC sampling, on probabilistic inference benchmarks, deep learning applications, and a novel decision-making application in network design for wildlife conservation. I also apply the embedding technique to automated reasoning and machine learning for dimensionality reduction in scientific discovery. As one example, I propose the use of embeddings based on Fourier analysis as a compact representation of high-dimensional probability distributions. As a second example,  I show how human computation, crowdsourcing, and parallel computation can identify key backdoor information, thereby drastically reducing the computation time from days to minutes in a dimensionality reduction application with complex physical constraints. Our novel integration of reasoning and learning has led to the discovery of new solar light absorbers by solving a dimensionality reduction problem to characterize the crystal structure of metal oxide materials using X-ray diffraction data.</dim:field>
   <dim:field mdschema="dc" element="language" qualifier="iso">en_US</dim:field>
   <dim:field mdschema="dc" element="rights" lang="*">Attribution-NonCommercial 4.0 International</dim:field>
   <dim:field mdschema="dc" element="rights" qualifier="uri" lang="*">https://creativecommons.org/licenses/by-nc/4.0/</dim:field>
   <dim:field mdschema="dc" element="subject">Computer science</dim:field>
   <dim:field mdschema="dc" element="title">Combining Reasoning and Learning for Multi-stage Inference in Computational Sustainability and Scientific Discovery</dim:field>
   <dim:field mdschema="dc" element="type">dissertation or thesis</dim:field>
   <dim:field mdschema="dc" element="format" qualifier="mimetype">application/pdf</dim:field>
   <dim:field mdschema="thesis" element="degree" qualifier="discipline">Computer Science</dim:field>
   <dim:field mdschema="thesis" element="degree" qualifier="grantor">Cornell University</dim:field>
   <dim:field mdschema="thesis" element="degree" qualifier="level">Doctor of Philosophy</dim:field>
   <dim:field mdschema="thesis" element="degree" qualifier="name">Ph. D., Computer Science</dim:field>
   <dim:field mdschema="dcterms" element="license">https://hdl.handle.net/1813/59810</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">Xue, Yexiang</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="d9705c3a-6dce-4ae3-ac8e-a09dc047cbb8">
	&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>Combining Reasoning and Learning for Multi-stage Inference in Computational Sustainability and Scientific Discovery&lt;/Title>
   	&lt;PublishedIn>
    	&lt;Publication>
      	&lt;/Publication>
   	&lt;/PublishedIn>
   	&lt;PublicationDate>2018-05-30&lt;/PublicationDate>
   	&lt;DOI>https://doi.org/10.7298/X4V986B4&lt;/DOI>
   	&lt;Authors>
      	&lt;Author>
        	&lt;DisplayName>Xue, Yexiang&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;License>https://creativecommons.org/licenses/by-nc/4.0/&lt;/License>
    &lt;Keyword>Computer science&lt;/Keyword>
   	&lt;Abstract>Problems at the intersection of reasoning, optimization, and learning often involve multi-stage inference. For example, making decisions based on machine learning models often leads to multi-stage inference problems where probabilistic models learned from data are embedded as first-stage subproblems within a global second-stage problem for decision-making. Multi-agent reasoning also involves multi-stage inference, since the reasoning of any given agent has to incorporate the goals of the other agents. As a result, the decision processes of the other agents are embedded as first-stage subproblems within the overall decision-making problem of that agent. With formal complexities beyond NP, multi-stage inference problems are often highly intractable. In this thesis, I introduce a novel computational framework, based on embeddings, to tackle multi-stage inference problems. Our embedding technique approximates the intractable sub-problems of a multi-stage inference problem through a series of novel representations. We then embed these representations into the global problem, effectively reducing a multi-stage inference to a single-stage inference. As one example, I present a novel way to encode the reward allocation problem for a two-stage organizer--agent game as a single-stage optimization. The encoding embeds an approximation of the agents&amp;apos; decision-making process into the organizer&amp;apos;s problem in the form of linear constraints. We apply this methodology to eBird, a well-established citizen-science program for collecting bird observations, in a game called Avicaching. Our AI-based reward allocation was shown to be highly effective, surpassing the expectations of the eBird organizers and bird conservation experts. As another example, I present a novel constant approximation algorithm to solve stochastic optimization problems which identify the optimal policy that maximizes the expectation of a stochastic objective. To tackle this problem, I propose the embedding of its intractable counting subproblems as queries to NP oracles subject to additional XOR constraints. As a result, the entire problem is encoded as a single NP-equivalent optimization. This approach outperforms state-of-the-art solvers based on variational inference and MCMC sampling, on probabilistic inference benchmarks, deep learning applications, and a novel decision-making application in network design for wildlife conservation. I also apply the embedding technique to automated reasoning and machine learning for dimensionality reduction in scientific discovery. As one example, I propose the use of embeddings based on Fourier analysis as a compact representation of high-dimensional probability distributions. As a second example,  I show how human computation, crowdsourcing, and parallel computation can identify key backdoor information, thereby drastically reducing the computation time from days to minutes in a dimensionality reduction application with complex physical constraints. Our novel integration of reasoning and learning has led to the discovery of new solar light absorbers by solving a dimensionality reduction problem to characterize the crystal structure of metal oxide materials using X-ray diffraction data.&lt;/Abstract>
	&lt;Access xmlns="http://purl.org/coar/access_right" 
    >
    &lt;/Access>
&lt;/Publication>
</dim:field>
</dim:dim>
</metadata></record></GetRecord></OAI-PMH>