Bayesian analysis of online newspaper log data

Wettig, Hannes; Lahtinen, Jussi; Lepola, Tuomas; Myllymäki, Petri; Tirri, Henry

doi:10.1109/saintw.2003.1210173

Cited by 8 publications

(10 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Combining a memory-based technique with a model-based part has been suggested (Yu et al 2004) to avoid the problem of new users not having enough ratings for reliable predictions. Recent work includes probabilistic and information-theoretic models, for instance (Hofmann 2004;Jin and Si 2004;Wettig et al 2003;Zitnick and Kanade 2004). An interesting family of models are the latent component models, which have been successfully used in document modeling but also in collaborative filtering Erosheva et al 2004;Hofmann 2004;Keller and Bengio 2004;Marlin 2004;Marlin and Zemel 2004;McCallum et al 2004;Popescul et al 2001;Pritchard et al 2000;Rosen-Zvi et al 2004;Si and Jin 2003;Yu et al 2005aYu et al , 2005b.…”

Section: Introductionmentioning

confidence: 99%

“…Finally, although our model is not far from the diverse set of so-called biclustering models (Madeira and Oliveira 2004;Tanay et al 2006;Wettig et al 2003), we aim at prediction instead of clustering, and therefore it is enough that the latent structure makes the predictions accurate. Because the model is computed by sampling, we do not obtain a single explicit set of clusters.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Latent grouping models for user preference prediction

2008

View full text Add to dashboard Cite

We tackle the problem of new users or documents in collaborative filtering. Generalization over users by grouping them into user groups is beneficial when a rating is to be predicted for a relatively new document having only few observed ratings. Analogously, generalization over documents improves predictions in the case of new users. We show that if either users and documents or both are new, two-way generalization becomes necessary. We demonstrate the benefits of grouping of users, grouping of documents, and two-way grouping, with artificial data and in two case studies with real data. We have introduced a probabilistic latent grouping model for predicting the relevance of a document to a user. The model assumes a latent group structure for both users and items. We compare the model against a state-of-the-art method, the User Rating Profile model, where only the users have a latent group structure. We compute the posterior of both models by Gibbs sampling. The Two-Way Model predicts relevance more accurately when the target consists of both new documents and new users. The reason is that generalization over documents becomes beneficial for new documents and at the same time generalization over users is needed for new users.

show abstract

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

Latent grouping models for user preference prediction

2008

View full text Add to dashboard Cite

show abstract

“…• Collaborative filtering, either memory-based [4] or model-based [18], which allows matching of an agent onto an agent coalition. A corresponding task can be allocated to alleviate an incident on the basis of the similarity between the coalition formation profile of an agent and those of other agents.…”

Section: Introductionmentioning

confidence: 99%

Self-organizing mobile surveillance security networks

Ferro

Salden

2007

2007 2nd Bio-Inspired Models of Network, Information and Computing Systems

View full text Add to dashboard Cite

We develop and evaluate ASK-ASSIST, a decision support system that enables automatically setting up communication networks and sharing knowledge amongst security personnel, in particular when they are confronted with a series of unexpected incidents while on patrol. Currently, team leaders of mobile security guards are still forced to assess the consequences of a series of incidents and to take decisions in a heuristic way. On the basis of incomplete and uncertain information they have to determine the security situation, who to bring in contact with whom, and which personnel or which information systems should be lined up in order to resolve the incidents as a whole. ASK-ASSIST, however, automates robust dynamic formation and coordination of the right coalitions of security personnel, such that they support human improvisation in determining the nature of incidents and taking decisions by the coalitions. These coalitions comprise security personnel, communication security infrastructures and knowledge management security infrastructures. ASK-ASSIST is based on a framework for selforganization that generates ranked lists of incident-specific critical (re)configurations of coalitions in which joint tasks are assigned to agents associated with roles, like guards or team leaders. Our main contributions are the grounding and evaluation of the framework underlying ASK-ASSIST. We instantiate, validate and test the system on the basis of real data. These data concern organizational structures reflected in the coalitions and implicit or explicit feedback provided by personnel from a private security company. The provided feedback relates to the daily operational mobile security surveillance processes including those that involve functionalities of ASK-ASSIST.

show abstract

“…Model-based approaches are justified by the poor scaling of the memory-based techniques. Recent work includes probabilistic and informationtheoretic models, see for instance [3,4,5,6].…”

mentioning

confidence: 99%

Two-Way Grouping by One-Way Topic Models

Savia

Puolamäki

Kaski

2009

Advances in Intelligent Data Analysis VIII

View full text Add to dashboard Cite

Abstract. We tackle the problem of new users or documents in collaborative filtering. Generalization over users by grouping them into user groups is beneficial when a rating is to be predicted for a relatively new document having only few observed ratings. The same applies for documents in the case of new users. We have shown earlier that if there are both new users and new documents, two-way generalization becomes necessary, and introduced a probabilistic Two-Way Model for the task. The task of finding a two-way grouping is a non-trivial combinatorial problem, which makes it computationally difficult. We suggest approximating the Two-Way Model with two URP models; one that groups users and one that groups documents. Their two predictions are combined using a product of experts model. This combination of two one-way models achieves even better prediction performance than the original Two-Way Model.

show abstract

Bayesian analysis of online newspaper log data

Abstract: Abstract

Cited by 8 publications

References 4 publications

Latent grouping models for user preference prediction

Latent grouping models for user preference prediction

Self-organizing mobile surveillance security networks

Two-Way Grouping by One-Way Topic Models

Contact Info

Product

Resources

About