A Novel Way of Computing Similarities between Nodes of a Graph, with Application to Collaborative Recommendation

Fouss, François; Pirotte, Alain; Saerens, Marco

doi:10.1109/wi.2005.9

Cited by 60 publications

(87 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…P 3 and P 5 perform random walks of fixed length 3 and 5, respectively, starting from a target user vertex. P 3 α , which raises the transition probabilities to the power of α, is more accurate than the methods proposed in [Fouss et al 2005] and [Gori et al 2007]. They also show that approximations obtained using random walk sampling are more efficient and scalable compared to methods based on matrix calculations.…”

Section: Related Workmentioning

confidence: 95%

“…The use of a graph-based model for recommendations was first introduced in [Aggarwal et al 1999]. To apply a bipartite user-item-feedback graph G was proposed in [Huang et al 2004] and several projects [Baluja et al 2008;Bogers 2010;Cooper et al 2014;Fouss et al 2005;Gori et al 2007;Jamali and Ester 2009;Lee et al 2012;Xiang et al 2010] extended this approach. We classify them as vertex ranking algorithms because their main idea is to rank the vertices in the graph based on their similarities with the target user and use the ranking to generate recommendations.…”

Section: Related Workmentioning

confidence: 99%

“…Fouss et al [Fouss et al 2005] introduced the idea of using random walks on G to rank the vertices. Vertices are ranked or scored based on quantities like hitting time, average commute time or the entries in the Moore-Penrose pseudo inverse of the Laplacian matrix of the graph (L + ).…”

Section: Related Workmentioning

confidence: 99%

“…These algorithms are more accurate than previous methods (e.g., [Fouss et al 2005]) with the additional benefit of being computationally efficient and scalable.…”

Section: Introductionmentioning

confidence: 99%

“…We show using three datasets (two public, one enterprise) that RP 3 β augments long-tail item recommendations while keeping accuracy high. Second, we empirically compare the performance of vertex ranking algorithms [Fouss et al 2005;Zhou et al 2010;Cooper et al 2014] including our own RP 3 β with traditional state-of-the-art methods. We find that some vertex ranking algorithms achieve comparable or better performance than the traditional ones.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Updatable, Accurate, Diverse, and Scalable Recommendations for Interactive Applications

Paudel

Christoffel

Newell

et al. 2016

ACM Trans. Interact. Intell. Syst.

View full text Add to dashboard Cite

Recommender systems form the backbone of many interactive systems. They incorporate user feedback to personalize the user experience typically via personalized recommendation lists. As users interact with a system, an increasing amount of data about a user's preferences becomes available, which can be leveraged for improving the systems' performance. Incorporating these new data into the underlying recommendation model is, however, not always straightforward. Many models used by recommender systems are computationally expensive and, therefore, have to perform offline computations to compile the recommendation lists. For interactive applications, it is desirable to be able to update the computed values as soon as new user interaction data is available: updating recommendations in interactive time using new feedback data leads to better accuracy and increases the attraction of the system to the users. Additionally, there is a growing consensus that accuracy alone is not enough and user satisfaction is also dependent on diverse recommendations. In this work, we tackle this problem of updating personalized recommendation lists for interactive applications in order to provide both accurate and diverse recommendations. To that end, we explore algorithms that exploit random walks as a sampling technique to obtain diverse recommendations without compromising on efficiency and accuracy. Specifically, we present a novel graph vertex ranking recommendation algorithm called RP3 that reranks items based on three-hop random walk transition probabilities. We show empirically that RP3 provides accurate recommendations with high long-tail item frequency at the top of the recommendation list. We also present approximate versions of RP3 and the two most accurate previously published vertex ranking algorithms based on random walk transition probabilities and show that these approximations converge with an increasing number of samples. To obtain interactively updatable recommendations, we additionally show how our algorithm can be extended for online updates at interactive speeds. The underlying random walk sampling technique makes it possible to perform the updates without having to recompute the values for the entire dataset. In an empirical evaluation with three real-world datasets, we show that RP3 provides highly accurate and diverse recommendations that can easily be updated with newly gathered information at interactive speeds ( 100ms). Recommender systems form the backbone of many interactive systems. They incorporate user feedback to personalize the user experience typically via personalized recommendation lists. As users interact with a system, an increasing amount of data about a user's preferences becomes available, which can be leveraged for improving the systems performance. Incorporating these new data into the underlying recommendation model is, however, not always straightforward. Many models used by recommender systems are computationally expensive and, therefore, have to perform offline computations to c...

show abstract

Section: Related Workmentioning

confidence: 95%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

“…These algorithms are more accurate than previous methods (e.g., [Fouss et al 2005]) with the additional benefit of being computationally efficient and scalable.…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Updatable, Accurate, Diverse, and Scalable Recommendations for Interactive Applications

Paudel

Christoffel

Newell

et al. 2016

ACM Trans. Interact. Intell. Syst.

View full text Add to dashboard Cite

show abstract

Community Detection in Biological Networks

Thakur

2011

Applied Statistics for Network Biology

View full text Add to dashboard Cite

Networks are everywhere and evolving in various forms, following certain characteristic patterns like small-world and scale-free characteristics. The basis of these forms is not the random addition of nodes and vertices to the system under study, but a systematic process of mutual evolution. This makes many aspects of such systems worth studying. Researchers have invested many years interpreting such systems, for instance how biological networks evolve and why humans form societies to name a few. Thus, seminal to the behavior to these systems are the way internal components are connected to each other. These connections breed structures that exemplify the nature and the behavior of the overall system. Advances in genomics have generated enormous amounts of data, currently gathered by biologists and now available to the scientific community at large. The categorization of this data naturally falls into various classes, ranging from protein sequences to complex molecules. Thus, challenges lie in understanding the properties and dynamics of such complex systems rather than gaining insight into the individual components. Fortunately, promises shown in statistical physics and network theory that offer a host of methods and tools to analyze large data can invariably be used in biology. Modeling interactions between proteins, cells, and other genomic data as network graphs and applying statistical techniques on them can greatly benefit us in scrutinizing the intricacies that exist in these datasets. These methods not only simplify the representations, but also aid in searching for general principles of organization and evolution of biological structures. For example, smallworld analysis is used to explain protein-protein interaction networks and scale-free invariance properties exist for in silico evolution of metabolic networks, which encode an increasing amount of information about their environment. The questions often arises as to which properties of such systems are worth studying and how are the emergent behaviors actually affecting the overall functionality of such systems. Visualization is a potential tool that was used as a first step to display the connectivity patterns between various components and describe the structure as a whole. Another

show abstract

Fuzzy Models for Link Prediction in Social Networks

Bastani

Jafarabad

Zarandi

2013

Int. J. Intell. Syst.

View full text Add to dashboard Cite

Predicting missing links and links that may occur in the future in social networks is an attention grabbing topic amid the social network analysts. Owing to the relationship between human‐based system and social sciences in this field, granular computing can help us to model the systems more effectively. The present study aims to propose two new similarity indices, based on granular computing approach and fuzzy logic. It also presents a new hybrid model for creating synergy between various link prediction models. Results show that fuzzy system analysis, in comparison with the crisp approach, can make more effective predictions through better expression of network characteristics. The indices are tested on collaboration networks. It is found that the accuracy of predictions is significantly higher than the crisp approach. It can modify the models for computing the strength of the links and/or predicting the evolutions of the social networks.

show abstract

A Novel Way of Computing Similarities between Nodes of a Graph, with Application to Collaborative Recommendation

Cited by 60 publications

References 8 publications

Updatable, Accurate, Diverse, and Scalable Recommendations for Interactive Applications

Updatable, Accurate, Diverse, and Scalable Recommendations for Interactive Applications

Community Detection in Biological Networks

Fuzzy Models for Link Prediction in Social Networks

Contact Info

Product

Resources

About