An Infinite Latent Attribute Model for Network Data. Palla, K., Knowles, D. A., & Ghahramani, Z. In *29th International Conference on Machine Learning (ICML 2012)*, pages 1607–1614, 2012. Paper abstract bibtex Latent variable models for network data extract a summary of the relational structure underlying an observed network. The simplest possible models subdivide nodes of the network into clusters; the probability of a link between any two nodes then depends only on their cluster assignment. Currently available models can be classified by whether clusters are disjoint or are allowed to overlap. These models can explain clustering structure. Hierarchical Bayesian models provide a natural approach to capture more complex dependencies. We propose a model in which objects are characterised by a latent feature vector. Each feature is itself partitioned into disjoint groups (subclusters), corresponding to a second layer of hierarchy. In experimental comparisons, the model achieves significantly improved predictive performance on social and biological link prediction tasks. The results indicate that models with a single layer hierarchy over-simplify real networks.

@inproceedings{palla2012infinite,
Abstract = {Latent variable models for network data extract
a summary of the relational structure
underlying an observed network. The simplest
possible models subdivide nodes of the
network into clusters; the probability of a link
between any two nodes then depends only on
their cluster assignment. Currently available
models can be classified by whether clusters
are disjoint or are allowed to overlap. These
models can explain clustering structure.
Hierarchical Bayesian models provide
a natural approach to capture more complex
dependencies. We propose a model in which
objects are characterised by a latent feature
vector. Each feature is itself partitioned into
disjoint groups (subclusters), corresponding
to a second layer of hierarchy. In experimental
comparisons, the model achieves significantly
improved predictive performance on
social and biological link prediction tasks.
The results indicate that models with a single
layer hierarchy over-simplify real networks.},
Author = {Palla, Konstantina and Knowles, David A. and Ghahramani, Zoubin},
Booktitle = {29th International Conference on Machine Learning (ICML 2012)},
Keywords = {Machine Learning/Statistics},
Pages = {1607--1614},
Title = {{An Infinite Latent Attribute Model for Network Data}},
Url = {http://icml.cc/2012/papers/785.pdf},
Year = {2012}}

Downloads: 0

{"_id":"9noAQDGiwbrxGGehH","bibbaseid":"palla-knowles-ghahramani-aninfinitelatentattributemodelfornetworkdata-2012","downloads":0,"creationDate":"2017-05-18T02:00:25.680Z","title":"An Infinite Latent Attribute Model for Network Data","author_short":["Palla, K.","Knowles, D. A.","Ghahramani, Z."],"year":2012,"bibtype":"inproceedings","biburl":"http://cs.stanford.edu/people/davidknowles/reordered.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","abstract":"Latent variable models for network data extract a summary of the relational structure underlying an observed network. The simplest possible models subdivide nodes of the network into clusters; the probability of a link between any two nodes then depends only on their cluster assignment. Currently available models can be classified by whether clusters are disjoint or are allowed to overlap. These models can explain clustering structure. Hierarchical Bayesian models provide a natural approach to capture more complex dependencies. We propose a model in which objects are characterised by a latent feature vector. Each feature is itself partitioned into disjoint groups (subclusters), corresponding to a second layer of hierarchy. In experimental comparisons, the model achieves significantly improved predictive performance on social and biological link prediction tasks. The results indicate that models with a single layer hierarchy over-simplify real networks.","author":[{"propositions":[],"lastnames":["Palla"],"firstnames":["Konstantina"],"suffixes":[]},{"propositions":[],"lastnames":["Knowles"],"firstnames":["David","A."],"suffixes":[]},{"propositions":[],"lastnames":["Ghahramani"],"firstnames":["Zoubin"],"suffixes":[]}],"booktitle":"29th International Conference on Machine Learning (ICML 2012)","keywords":"Machine Learning/Statistics","pages":"1607–1614","title":"An Infinite Latent Attribute Model for Network Data","url":"http://icml.cc/2012/papers/785.pdf","year":"2012","bibtex":"@inproceedings{palla2012infinite,\n\tAbstract = {Latent variable models for network data extract\na summary of the relational structure\nunderlying an observed network. The simplest\npossible models subdivide nodes of the\nnetwork into clusters; the probability of a link\nbetween any two nodes then depends only on\ntheir cluster assignment. Currently available\nmodels can be classified by whether clusters\nare disjoint or are allowed to overlap. These\nmodels can explain clustering structure.\nHierarchical Bayesian models provide\na natural approach to capture more complex\ndependencies. We propose a model in which\nobjects are characterised by a latent feature\nvector. Each feature is itself partitioned into\ndisjoint groups (subclusters), corresponding\nto a second layer of hierarchy. In experimental\ncomparisons, the model achieves significantly\nimproved predictive performance on\nsocial and biological link prediction tasks.\nThe results indicate that models with a single\nlayer hierarchy over-simplify real networks.},\n\tAuthor = {Palla, Konstantina and Knowles, David A. and Ghahramani, Zoubin},\n\tBooktitle = {29th International Conference on Machine Learning (ICML 2012)},\n\tKeywords = {Machine Learning/Statistics},\n\tPages = {1607--1614},\n\tTitle = {{An Infinite Latent Attribute Model for Network Data}},\n\tUrl = {http://icml.cc/2012/papers/785.pdf},\n\tYear = {2012}}\n\n","author_short":["Palla, K.","Knowles, D. A.","Ghahramani, Z."],"key":"palla2012infinite","id":"palla2012infinite","bibbaseid":"palla-knowles-ghahramani-aninfinitelatentattributemodelfornetworkdata-2012","role":"author","urls":{"Paper":"http://icml.cc/2012/papers/785.pdf"},"keyword":["Machine Learning/Statistics"],"downloads":0},"search_terms":["infinite","latent","attribute","model","network","data","palla","knowles","ghahramani"],"keywords":["machine learning","statistics","biology","genetics","machine learning/statistics"],"authorIDs":[],"dataSources":["E5kTWRuqMhy8QxJbW"]}