Collaborative personalized top-k processing

Bai, Xiao; Guerraoui, Rachid; Kermarrec, Anne-Marie; Leroy, Vincent

doi:10.1145/2043652.2043659

Cited by 12 publications

(18 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similarly to KPS and previous gossipbased protocols [6,7,18,24], NNDescent relies on two principles: (i) the assumption that "The neighbor of a neighbor is also likely to be a neighbor" and (ii) a reduced candidate set to limit the number of similarity computations at each iteration.…”

Section: Nndescent Algorithmmentioning

confidence: 99%

“…In such systems, each peer is connected to a subset of other peers in the network and periodically exchanges some information with one of its neighbors. While such protocols have been initially used to build uniform random topologies [19,21], they have also been applied in the context of several applications to cluster peers according to some specific metric (interest, overlap, etc) to build networks of arbitrary structure [18,24] or to support various applications such as query expansion [7], top-k queries [6] or news recommendation [9]. In such a system, the use of random nodes ensures that connectivity is maintained, each node is responsible to discover its KNN nodes by periodically exchanging neighborhood information with other peers.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Nearest Neighbors Graph Construction: Peer Sampling to the Rescue

2016

Self Cite

View full text Add to dashboard Cite

Abstract. In this paper, we propose an efficient KNN service, called KPS (KNN-Peer-Sampling). The KPS service can be used in various contexts e.g. recommendation systems, information retrieval and data mining. KPS borrows concepts from P2P gossip-based clustering protocols to provide a localized and efficient KNN computation in large-scale systems. KPS is a sampling-based iterative approach, combining randomness, to provide serendipity and avoid local minimum, and clustering, to ensure fast convergence. We compare KPS against the state of the art KNN centralized computation algorithm NNDescent, on multiple datasets. The experiments confirm the efficiency of KPS over NNDescent: KPS improves significantly on the computational cost while converging quickly to a close to optimal KNN graph. For instance, the cost, expressed in number of pairwise similarity computations, is reduced by ≈ 23% and ≈ 49% to construct high quality KNN graphs for Jester and MovieLens datasets, respectively. In addition, the randomized nature of KPS ensures eventual convergence, not always achieved with NNDescent.

show abstract

Section: Nndescent Algorithmmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Nearest Neighbors Graph Construction: Peer Sampling to the Rescue

2016

Self Cite

View full text Add to dashboard Cite

show abstract

“…As a consequence, each user's cluster may still carry redundant user profiles, because there is no explicit diversification. In [1], Bai et al propose a solution for personalized P2P top-k search in the context of collaborative tagging systems, called P4Q. In this solution, the users are clustered based on relevance through gossip protocols.…”

Section: Related Workmentioning

confidence: 99%

“…In this context of large scale distribution of users and data, a general solution to data sharing is offered by distributed search and recommendation [1,2]. In this paper, we adopt a peer-to-peer gossip-based approach, because it provides important properties such as scalability, dynamicity, autonomy and decentralized control.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Exploiting Diversification in Gossip-Based Recommendation

Servajean

Pacitti

Liroz-Gistau

et al. 2014

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In the context of Web 2.0, the users become massive producers of diverse data that can be stored in a large variety of systems. The fact that the users' data spaces are distributed in many different systems makes data sharing difficult. In this context of large scale distribution of users and data, a general solution to data sharing is offered by distributed search and recommendation. In particular, gossip-based approaches provide scalability, dynamicity, autonomy and decentralized control. Generally, in gossip-based search and recommendation, each user constructs a cluster of "relevant" users that will be employed in the processing of queries. However, considering only relevance introduces a significant amount of redundancy among users. As a result, when a query is submitted, as the user profiles in each user's cluster are quite similar, the probability of retrieving the same set of relevant items increases, and recall results are limited. In this paper, we propose a gossip-based search and recommendation approach that is based on a new clustering score, called usefulness, that combines relevance and diversity, and we present the corresponding new gossip-based clustering algorithm. We validate our proposal with an experimental evaluation using three datasets based on MovieLens, Flickr and LastFM. Compared with state of the art solutions, we obtain major gains with a three order of magnitude recall improvement when using the notion of usefulness regardless of the relevance score between two users used.

show abstract

Increasing Coverage in Distributed Search and Recommendation with Profile Diversity

Servajean

Pacitti

Liroz-Gistau

et al. 2015

Transactions on Large-Scale Data- And Knowledge-Centered Systems XXII

View full text Add to dashboard Cite

International audienceWith the advent of Web 2.0 users are producing bigger and bigger amounts of diverse data, which are stored in a large variety of systems. Since the users’ data spaces are scattered among those independent systems, data sharing becomes a challenging problem. Distributed search and recommendation provides a general solution for data sharing and among its various alternatives, gossip-based approaches are particularly interesting as they provide scalability, dynamicity, autonomy and decentralized control. Generally, in these approaches each participant maintains a cluster of “relevant” users, which are later employed in query processing. However, as we show in the paper, only considering relevance in the construction of the cluster introduces a significant amount of redundancy among users, which in turn leads to reduced recall. Indeed, when a query is submitted, due to the high similarity among the users in a cluster, the probability of retrieving the same set of relevant items increases, thus limiting the amount of distinct results that can be obtained. In this paper, we propose a gossip-based search and recommendation approach that is based on diversity-based clustering scores. We present the resultant new gossip-based clustering algorithms and validate them through experimental evaluation over four real datasets, based on MovieLens-small, MovieLens, LastFM and Delicious. Compared with state of the art solutions, we show that taking into account diversity-based clustering score enables to obtain major gains in terms of recall while reducing the number of users involved during query processing

show abstract

Collaborative personalized top-k processing

Cited by 12 publications

References 25 publications

Nearest Neighbors Graph Construction: Peer Sampling to the Rescue

Nearest Neighbors Graph Construction: Peer Sampling to the Rescue

Exploiting Diversification in Gossip-Based Recommendation

Increasing Coverage in Distributed Search and Recommendation with Profile Diversity

Contact Info

Product

Resources

About