{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/95281"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/95281","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Distributed content collection and rank aggregation","abstract":"Despite the substantial literature on recommendation systems, there have been few studies in distributed settings, where peers provide recommendations locally. Motivated by word of mouth type of social behavior and the advantages of sharing resources, we analyze an online distributed recommendation system with joint content collection and rank aggregation. In such a system, peers contact each other and exchange partial preference information about items, which, for example, could be videos. Peers use recommendation strategies to make decisions with limited knowledge and collect items that are available from the contacted peers. The goal is to maximize the rate at which peers collect their most preferred items. Correlated preferences are modeled as rankings generated by a Plackett-Luce ranking model with Zipf popularity distribution. We establish a performance upper bound and use intuition provided by the bound to design recommendation strategies with a range of complexity. Among these, the direct recommendation rule emerges as being particularly simple and yet effective. The direct recommendation rule is found to be remarkably robust, working well over a broad range of correlation of preferences, initial video availability, storage size, peer arrival pattern, and performance metric. Correlated preferences are modeled as scores generated using an independent crossover model. In order to explore performance for large scale networks, we identify the fluid limit as the number of videos goes to infinity for a mean field limit derived for the number of peers going to infinity under a direct recommendation rule. Simulation results show that the limit analysis accurately predicts performance, not only for the independent crossover model with scores, but also a model with rankings. The performance of the direct recommendation rule is shown to be near optimal for large scale systems. Correlated preferences are modeled as scores generated using a two-stage independent crossover model. We propose four recommendation strategies for heterogeneous preferences. We find that a simple rule, called the nearest stored preference rule, is as effective as the more complex rules. The performance of all the rules is far from a performance upper bound in case the peers in different clusters are nearly independent. We find through simulation that the gap can be nearly closed by using either exponential accumulation of information or neighbor assignments such that most neighbors have similar preferences.","abstract_html":"Despite the substantial literature on recommendation systems, there have been few studies in distributed settings, where peers provide recommendations locally. Motivated by word of mouth type of social behavior and the advantages of sharing resources, we analyze an online distributed recommendation system with joint content collection and rank aggregation. In such a system, peers contact each other and exchange partial preference information about items, which, for example, could be videos. Peers use recommendation strategies to make decisions with limited knowledge and collect items that are available from the contacted peers. The goal is to maximize the rate at which peers collect their most preferred items. Correlated preferences are modeled as rankings generated by a Plackett-Luce ranking model with Zipf popularity distribution. We establish a performance upper bound and use intuition provided by the bound to design recommendation strategies with a range of complexity. Among these, the direct recommendation rule emerges as being particularly simple and yet effective. The direct recommendation rule is found to be remarkably robust, working well over a broad range of correlation of preferences, initial video availability, storage size, peer arrival pattern, and performance metric. Correlated preferences are modeled as scores generated using an independent crossover model. In order to explore performance for large scale networks, we identify the fluid limit as the number of videos goes to infinity for a mean field limit derived for the number of peers going to infinity under a direct recommendation rule. Simulation results show that the limit analysis accurately predicts performance, not only for the independent crossover model with scores, but also a model with rankings. The performance of the direct recommendation rule is shown to be near optimal for large scale systems. Correlated preferences are modeled as scores generated using a two-stage independent crossover model. We propose four recommendation strategies for heterogeneous preferences. We find that a simple rule, called the nearest stored preference rule, is as effective as the more complex rules. The performance of all the rules is far from a performance upper bound in case the peers in different clusters are nearly independent. We find through simulation that the gap can be nearly closed by using either exponential accumulation of information or neighbor assignments such that most neighbors have similar preferences.","abstract_has_math":false,"creators":["Yang, James Yifei"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"Ph.D.","degree_level":"Dissertation","degree_discipline":"Electrical & Computer Engr","degree_department":null,"school":null,"contributors":["Hajek, Bruce","Srikant, Rayadurgam","Vaidya, Nitin","Oh, Sewoong","Chiu, Dah Ming"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2017,"date_issued":"2017-03-01T15:46:02Z","date_published":"2017-03-01T15:46:02Z","updated_at":"2026-07-22T22:26:35Z","subjects":["Rank Aggregation","Score Aggregation","Mean Field","Scaling","Recommendation System","Peer-to-Peer","Content Collection","Independent Crossover Model","Plackett-Luce","Zipf","Clustering","Multi-Cluster","Hypothesis Testing"],"languages":["en"],"rights":["Copyright 2016 James Yang"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/2142/95281","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Hajek, Bruce","Srikant, Rayadurgam","Vaidya, Nitin","Oh, Sewoong","Chiu, Dah Ming"]},{"key":"dc:creator","label":"Author","values":["Yang, James Yifei"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2017-03-01T15:46:02Z","2016-09-13","2016-12"]},{"key":"dc:type","label":"Dc Type","values":["text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical & Computer Engr"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Rank Aggregation","Score Aggregation","Mean Field","Scaling","Recommendation System","Peer-to-Peer","Content Collection","Independent Crossover Model","Plackett-Luce","Zipf","Clustering","Multi-Cluster","Hypothesis Testing"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2016 James Yang"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/2142/95281"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Despite the substantial literature on recommendation systems, there have been few studies in distributed settings, where peers provide recommendations locally. Motivated by word of mouth type of social behavior and the advantages of sharing resources, we analyze an online distributed recommendation system with joint content collection and rank aggregation. In such a system, peers contact each other and exchange partial preference information about items, which, for example, could be videos. Peers use recommendation strategies to make decisions with limited knowledge and collect items that are available from the contacted peers. The goal is to maximize the rate at which peers collect their most preferred items. Correlated preferences are modeled as rankings generated by a Plackett-Luce ranking model with Zipf popularity distribution. We establish a performance upper bound and use intuition provided by the bound to design recommendation strategies with a range of complexity. Among these, the direct recommendation rule emerges as being particularly simple and yet effective. The direct recommendation rule is found to be remarkably robust, working well over a broad range of correlation of preferences, initial video availability, storage size, peer arrival pattern, and performance metric. Correlated preferences are modeled as scores generated using an independent crossover model. In order to explore performance for large scale networks, we identify the fluid limit as the number of videos goes to infinity for a mean field limit derived for the number of peers going to infinity under a direct recommendation rule. Simulation results show that the limit analysis accurately predicts performance, not only for the independent crossover model with scores, but also a model with rankings. The performance of the direct recommendation rule is shown to be near optimal for large scale systems. Correlated preferences are modeled as scores generated using a two-stage independent crossover model. We propose four recommendation strategies for heterogeneous preferences. We find that a simple rule, called the nearest stored preference rule, is as effective as the more complex rules. The performance of all the rules is far from a performance upper bound in case the peers in different clusters are nearly independent. We find through simulation that the gap can be nearly closed by using either exponential accumulation of information or neighbor assignments such that most neighbors have similar preferences.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2017-02-28 without embargo terms","The student, James Yang, accepted the attached license on 2016-09-12 at 21:18.","The student, James Yang, submitted this Dissertation for approval on 2016-09-12 at 21:54.","This Dissertation was approved for publication on 2016-09-13 at 14:00.","DSpace SAF Submission Ingestion Package generated from Vireo submission #10154 on 2017-02-28 at 14:45:58","Made available in DSpace on 2017-03-01T15:46:02Z (GMT). No. of bitstreams: 3 YANG-DISSERTATION-2016.pdf: 1241425 bytes, checksum: 6aba4edd0283f8614d2ff7bf6442f3a4 (MD5) LICENSE.txt: 4207 bytes, checksum: 8434caa6d3bed2bcef060044e7f12461 (MD5) PROQUEST_LICENSE.txt: 4553 bytes, checksum: 68b3f7a45d0c310132241f69fed5f73b (MD5) Previous issue date: 2016-09-13"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Distributed content collection and rank aggregation"]}]}],"canonical_facts":{"dc:contributor":["Hajek, Bruce","Srikant, Rayadurgam","Vaidya, Nitin","Oh, Sewoong","Chiu, Dah Ming"],"dc:creator":["Yang, James Yifei"],"dc:date":["2017-03-01T15:46:02Z","2016-09-13","2016-12"],"dc:description":["Despite the substantial literature on recommendation systems, there have been few studies in distributed settings, where peers provide recommendations locally. Motivated by word of mouth type of social behavior and the advantages of sharing resources, we analyze an online distributed recommendation system with joint content collection and rank aggregation. In such a system, peers contact each other and exchange partial preference information about items, which, for example, could be videos. Peers use recommendation strategies to make decisions with limited knowledge and collect items that are available from the contacted peers. The goal is to maximize the rate at which peers collect their most preferred items. Correlated preferences are modeled as rankings generated by a Plackett-Luce ranking model with Zipf popularity distribution. We establish a performance upper bound and use intuition provided by the bound to design recommendation strategies with a range of complexity. Among these, the direct recommendation rule emerges as being particularly simple and yet effective. The direct recommendation rule is found to be remarkably robust, working well over a broad range of correlation of preferences, initial video availability, storage size, peer arrival pattern, and performance metric. Correlated preferences are modeled as scores generated using an independent crossover model. In order to explore performance for large scale networks, we identify the fluid limit as the number of videos goes to infinity for a mean field limit derived for the number of peers going to infinity under a direct recommendation rule. Simulation results show that the limit analysis accurately predicts performance, not only for the independent crossover model with scores, but also a model with rankings. The performance of the direct recommendation rule is shown to be near optimal for large scale systems. Correlated preferences are modeled as scores generated using a two-stage independent crossover model. We propose four recommendation strategies for heterogeneous preferences. We find that a simple rule, called the nearest stored preference rule, is as effective as the more complex rules. The performance of all the rules is far from a performance upper bound in case the peers in different clusters are nearly independent. We find through simulation that the gap can be nearly closed by using either exponential accumulation of information or neighbor assignments such that most neighbors have similar preferences.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2017-02-28 without embargo terms","The student, James Yang, accepted the attached license on 2016-09-12 at 21:18.","The student, James Yang, submitted this Dissertation for approval on 2016-09-12 at 21:54.","This Dissertation was approved for publication on 2016-09-13 at 14:00.","DSpace SAF Submission Ingestion Package generated from Vireo submission #10154 on 2017-02-28 at 14:45:58","Made available in DSpace on 2017-03-01T15:46:02Z (GMT). No. of bitstreams: 3 YANG-DISSERTATION-2016.pdf: 1241425 bytes, checksum: 6aba4edd0283f8614d2ff7bf6442f3a4 (MD5) LICENSE.txt: 4207 bytes, checksum: 8434caa6d3bed2bcef060044e7f12461 (MD5) PROQUEST_LICENSE.txt: 4553 bytes, checksum: 68b3f7a45d0c310132241f69fed5f73b (MD5) Previous issue date: 2016-09-13"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/2142/95281"],"dc:language":["en"],"dc:rights":["Copyright 2016 James Yang"],"dc:subject":["Rank Aggregation","Score Aggregation","Mean Field","Scaling","Recommendation System","Peer-to-Peer","Content Collection","Independent Crossover Model","Plackett-Luce","Zipf","Clustering","Multi-Cluster","Hypothesis Testing"],"dc:title":["Distributed content collection and rank aggregation"],"dc:type":["text"],"thesis:degree_discipline":["Electrical & Computer Engr"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Ph.D."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:26:35Z"}