Overlay management for fully distributed user-based collaborative filtering

Róbert Ormándi, István Hegedus, M. Jelasity

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Citations (Scopus)

Abstract

Offering personalized recommendation as a service in fully distributed applications such as file-sharing, distributed search, social networking, P2P television, etc, is an increasingly important problem. In such networked environments recommender algorithms should meet the same performance and reliability requirements as in centralized services. To achieve this is a challenge because a large amount of distributed data needs to be managed, and at the same time additional constraints need to be taken into account such as balancing resource usage over the network. In this paper we focus on a common component of many fully distributed recommender systems, namely the overlay network. We point out that the overlay topologies that are typically defined by node similarity have highly unbalanced degree distributions in a wide range of available benchmark datasets: a fact that has important-but so far largely overlooked-consequences on the load balancing of overlay protocols. We propose algorithms with a favorable convergence speed and prediction accuracy that also take load balancing into account. We perform extensive simulation experiments with the proposed algorithms, and compare them with known algorithms from related work on well-known benchmark datasets.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages446-457
Number of pages12
Volume6271 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2010
Event16th International Euro-Par Conference on Parallel Processing, Euro-Par 2010 - Ischia, Italy
Duration: Aug 31 2010Sep 3 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume6271 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other16th International Euro-Par Conference on Parallel Processing, Euro-Par 2010
CountryItaly
CityIschia
Period8/31/109/3/10

Fingerprint

Collaborative filtering
Collaborative Filtering
Overlay
Load Balancing
Resource allocation
Benchmark
Personalized Recommendation
Social Networking
Overlay networks
Overlay Networks
Recommender Systems
Recommender systems
Distributed Applications
Convergence Speed
Degree Distribution
Television
Balancing
Simulation Experiment
Distributed Systems
Sharing

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Ormándi, R., Hegedus, I., & Jelasity, M. (2010). Overlay management for fully distributed user-based collaborative filtering. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 6271 LNCS, pp. 446-457). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6271 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-15277-1_43

Overlay management for fully distributed user-based collaborative filtering. / Ormándi, Róbert; Hegedus, István; Jelasity, M.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6271 LNCS PART 1. ed. 2010. p. 446-457 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6271 LNCS, No. PART 1).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ormándi, R, Hegedus, I & Jelasity, M 2010, Overlay management for fully distributed user-based collaborative filtering. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 6271 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 6271 LNCS, pp. 446-457, 16th International Euro-Par Conference on Parallel Processing, Euro-Par 2010, Ischia, Italy, 8/31/10. https://doi.org/10.1007/978-3-642-15277-1_43
Ormándi R, Hegedus I, Jelasity M. Overlay management for fully distributed user-based collaborative filtering. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 6271 LNCS. 2010. p. 446-457. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-15277-1_43
Ormándi, Róbert ; Hegedus, István ; Jelasity, M. / Overlay management for fully distributed user-based collaborative filtering. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6271 LNCS PART 1. ed. 2010. pp. 446-457 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{5eec6c7d5f95492bbb786e490d2943d9,
title = "Overlay management for fully distributed user-based collaborative filtering",
abstract = "Offering personalized recommendation as a service in fully distributed applications such as file-sharing, distributed search, social networking, P2P television, etc, is an increasingly important problem. In such networked environments recommender algorithms should meet the same performance and reliability requirements as in centralized services. To achieve this is a challenge because a large amount of distributed data needs to be managed, and at the same time additional constraints need to be taken into account such as balancing resource usage over the network. In this paper we focus on a common component of many fully distributed recommender systems, namely the overlay network. We point out that the overlay topologies that are typically defined by node similarity have highly unbalanced degree distributions in a wide range of available benchmark datasets: a fact that has important-but so far largely overlooked-consequences on the load balancing of overlay protocols. We propose algorithms with a favorable convergence speed and prediction accuracy that also take load balancing into account. We perform extensive simulation experiments with the proposed algorithms, and compare them with known algorithms from related work on well-known benchmark datasets.",
author = "R{\'o}bert Orm{\'a}ndi and Istv{\'a}n Hegedus and M. Jelasity",
year = "2010",
doi = "10.1007/978-3-642-15277-1_43",
language = "English",
isbn = "3642152767",
volume = "6271 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "446--457",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - Overlay management for fully distributed user-based collaborative filtering

AU - Ormándi, Róbert

AU - Hegedus, István

AU - Jelasity, M.

PY - 2010

Y1 - 2010

N2 - Offering personalized recommendation as a service in fully distributed applications such as file-sharing, distributed search, social networking, P2P television, etc, is an increasingly important problem. In such networked environments recommender algorithms should meet the same performance and reliability requirements as in centralized services. To achieve this is a challenge because a large amount of distributed data needs to be managed, and at the same time additional constraints need to be taken into account such as balancing resource usage over the network. In this paper we focus on a common component of many fully distributed recommender systems, namely the overlay network. We point out that the overlay topologies that are typically defined by node similarity have highly unbalanced degree distributions in a wide range of available benchmark datasets: a fact that has important-but so far largely overlooked-consequences on the load balancing of overlay protocols. We propose algorithms with a favorable convergence speed and prediction accuracy that also take load balancing into account. We perform extensive simulation experiments with the proposed algorithms, and compare them with known algorithms from related work on well-known benchmark datasets.

AB - Offering personalized recommendation as a service in fully distributed applications such as file-sharing, distributed search, social networking, P2P television, etc, is an increasingly important problem. In such networked environments recommender algorithms should meet the same performance and reliability requirements as in centralized services. To achieve this is a challenge because a large amount of distributed data needs to be managed, and at the same time additional constraints need to be taken into account such as balancing resource usage over the network. In this paper we focus on a common component of many fully distributed recommender systems, namely the overlay network. We point out that the overlay topologies that are typically defined by node similarity have highly unbalanced degree distributions in a wide range of available benchmark datasets: a fact that has important-but so far largely overlooked-consequences on the load balancing of overlay protocols. We propose algorithms with a favorable convergence speed and prediction accuracy that also take load balancing into account. We perform extensive simulation experiments with the proposed algorithms, and compare them with known algorithms from related work on well-known benchmark datasets.

UR - http://www.scopus.com/inward/record.url?scp=78349232377&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78349232377&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15277-1_43

DO - 10.1007/978-3-642-15277-1_43

M3 - Conference contribution

AN - SCOPUS:78349232377

SN - 3642152767

SN - 9783642152764

VL - 6271 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 446

EP - 457

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -