@inproceedings{conf/pkdd/BalasubramanyanDC13, author = {Balasubramanyan, Ramnath and Dalvi, Bhavana Bharat and Cohen, William W.}, title = {From Topic Models to Semi-supervised Learning: Biasing Mixed-Membership Models to Exploit Topic-Indicative Features in Entity Clustering.}, editor = {Blockeel, Hendrik and Kersting, Kristian and Nijssen, Siegfried and Zelezný, Filip}, booktitle = {ECML/PKDD (2)}, series = {Lecture Notes in Computer Science}, publisher = {Springer}, year = {2013}, volume = {8189}, pages = {628-642}, url = {http://dblp.uni-trier.de/db/conf/pkdd/pkdd2013-2.html#BalasubramanyanDC13}, isbn = {978-3-642-40990-5}, keywords = {learning, models, sota, supervised, topic, toread} }

%0 = inproceedings %A = Balasubramanyan, Ramnath and Dalvi, Bhavana Bharat and Cohen, William W. %B = ECML/PKDD (2) %D = 2013 %I = Springer %T = From Topic Models to Semi-supervised Learning: Biasing Mixed-Membership Models to Exploit Topic-Indicative Features in Entity Clustering. %U = http://dblp.uni-trier.de/db/conf/pkdd/pkdd2013-2.html#BalasubramanyanDC13

Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open.</p> <p>We develop an algorithm based on <i>Supervised Random Walks</i> that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function.</p> <p>Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.

@inproceedings{backstrom2011supervised, author = {Backstrom, Lars and Leskovec, Jure}, title = {Supervised random walks: predicting and recommending links in social networks}, booktitle = {Proceedings of the fourth ACM international conference on Web search and data mining}, series = {WSDM '11}, publisher = {ACM}, address = {New York, NY, USA}, year = {2011}, pages = {635--644}, url = {http://doi.acm.org/10.1145/1935826.1935914}, doi = {10.1145/1935826.1935914}, isbn = {978-1-4503-0493-1}, keywords = {predicting, random, recommending, supervised, walks}, abstract = {Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open.</p> <p>We develop an algorithm based on <i>Supervised Random Walks</i> that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function.</p> <p>Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.} }

%0 = inproceedings %A = Backstrom, Lars and Leskovec, Jure %B = Proceedings of the fourth ACM international conference on Web search and data mining %C = New York, NY, USA %D = 2011 %I = ACM %T = Supervised random walks: predicting and recommending links in social networks %U = http://doi.acm.org/10.1145/1935826.1935914

Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open. We develop an algorithm based on Supervised Random Walks that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function. Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.

@misc{backstrom2010supervised, author = {Backstrom, L. and Leskovec, J.}, title = {Supervised Random Walks: Predicting and Recommending Links in Social Networks}, year = {2010}, note = {cite arxiv:1011.4071}, url = {http://arxiv.org/abs/1011.4071}, keywords = {predicting, random, recommending, supervised, toread, walks}, abstract = {Predicting the occurrence of links is a fundamental problem in networks. In the link prediction problem we are given a snapshot of a network and would like to infer which interactions among existing members are likely to occur in the near future or which existing interactions are we missing. Although this problem has been extensively studied, the challenge of how to effectively combine the information from the network structure with rich node and edge attribute data remains largely open. We develop an algorithm based on Supervised Random Walks that naturally combines the information from the network structure with node and edge level attributes. We achieve this by using these attributes to guide a random walk on the graph. We formulate a supervised learning task where the goal is to learn a function that assigns strengths to edges in the network such that a random walker is more likely to visit the nodes to which new links will be created in the future. We develop an efficient training algorithm to directly learn the edge strength estimation function. Our experiments on the Facebook social graph and large collaboration networks show that our approach outperforms state-of-the-art unsupervised approaches as well as approaches that are based on feature extraction.} }

%0 = misc %A = Backstrom, L. and Leskovec, J. %B = } %C = %D = 2010 %I = %T = Supervised Random Walks: Predicting and Recommending Links in Social Networks} %U = http://arxiv.org/abs/1011.4071

A significant portion of the world's text is tagged by readers on social bookmarking websites. Credit attribution is an inherent problem in these corpora because most pages have multiple tags, but the tags do not always apply with equal specificity across the whole document. Solving the credit attribution problem requires associating each word in a document with the most appropriate tags and vice versa. This paper introduces Labeled LDA, a topic model that constrains Latent Dirichlet Allocation by defining a one-to-one correspondence between LDA's latent topics and user tags. This allows Labeled LDA to directly learn word-tag correspondences. We demonstrate Labeled LDA's improved expressiveness over traditional LDA with visualizations of a corpus of tagged web pages from del.icio.us. Labeled LDA outperforms SVMs by more than 3 to 1 when extracting tag-specific document snippets. As a multi-label text classifier, our model is competitive with a discriminative baseline on a variety of datasets.

@inproceedings{Ramage:2009:LLS:1699510.1699543, author = {Ramage, Daniel and Hall, David and Nallapati, Ramesh and Manning, Christopher D.}, title = {Labeled LDA: A Supervised Topic Model for Credit Attribution in Multi-labeled Corpora}, booktitle = {Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1 - Volume 1}, series = {EMNLP '09}, publisher = {Association for Computational Linguistics}, address = {Stroudsburg, PA, USA}, year = {2009}, pages = {248--256}, url = {http://dl.acm.org/citation.cfm?id=1699510.1699543}, isbn = {978-1-932432-59-6}, keywords = {lda, model, supervised, topic, toread}, abstract = {A significant portion of the world's text is tagged by readers on social bookmarking websites. Credit attribution is an inherent problem in these corpora because most pages have multiple tags, but the tags do not always apply with equal specificity across the whole document. Solving the credit attribution problem requires associating each word in a document with the most appropriate tags and vice versa. This paper introduces Labeled LDA, a topic model that constrains Latent Dirichlet Allocation by defining a one-to-one correspondence between LDA's latent topics and user tags. This allows Labeled LDA to directly learn word-tag correspondences. We demonstrate Labeled LDA's improved expressiveness over traditional LDA with visualizations of a corpus of tagged web pages from del.icio.us. Labeled LDA outperforms SVMs by more than 3 to 1 when extracting tag-specific document snippets. As a multi-label text classifier, our model is competitive with a discriminative baseline on a variety of datasets.} }

%0 = inproceedings %A = Ramage, Daniel and Hall, David and Nallapati, Ramesh and Manning, Christopher D. %B = Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1 - Volume 1 %C = Stroudsburg, PA, USA %D = 2009 %I = Association for Computational Linguistics %T = Labeled LDA: A Supervised Topic Model for Credit Attribution in Multi-labeled Corpora %U = http://dl.acm.org/citation.cfm?id=1699510.1699543

@inproceedings{DBLP:conf/icdm/Schmidt-Thieme05, author = {Schmidt-Thieme, Lars}, title = {Compound Classification Models for Recommender Systems.}, booktitle = {Proceedings of the 5th IEEE International Conference on Data Mining (ICDM 2005), 27-30 November 2005}, publisher = {IEEE Computer Society}, address = {Houston, Texas, USA}, year = {2005}, pages = {378-385}, url = {http://doi.ieeecomputersociety.org/10.1109/ICDM.2005.46}, isbn = {0-7695-2278-5}, keywords = {recommender, learning, supervised} }

%0 = inproceedings %A = Schmidt-Thieme, Lars %B = Proceedings of the 5th IEEE International Conference on Data Mining (ICDM 2005), 27-30 November 2005 %C = Houston, Texas, USA %D = 2005 %I = IEEE Computer Society %T = Compound Classification Models for Recommender Systems. %U = http://doi.ieeecomputersociety.org/10.1109/ICDM.2005.46

@techreport{zhu05survey, author = {Zhu, Xiaojin}, title = {Semi-Supervised Learning Literature Survey}, year = {2005}, number = {1530}, url = {http://pages.cs.wisc.edu/~jerryzhu/pub/ssl_survey.pdf}, keywords = {clustering, learning, semi, supervised, toread} }

%0 = techreport %A = Zhu, Xiaojin %B = } %C = %D = 2005 %I = %T = Semi-Supervised Learning Literature Survey} %U = http://pages.cs.wisc.edu/~jerryzhu/pub/ssl_survey.pdf

Semi-supervised clustering uses a small amount of supervised
ta to aid unsupervised learning. One typical approach
ecifies a limited number of must-link and cannotlink
nstraints between pairs of examples. This paper
esents a pairwise constrained clustering framework and a
w method for actively selecting informative pairwise constraints
get improved clustering performance. The clustering
d active learning methods are both easily scalable
large datasets, and can handle very high dimensional data.
perimental and theoretical results confirm that this active
erying of pairwise constraints significantly improves the
curacy of clustering when given a relatively small amount
supervision.

@article{Basu:EtAl:04, author = {Basu, Sugato and Banerjee, Arindam and Mooney, Raymond J.}, title = {Active Semi-Supervision for Pairwise Constrained Clustering}, booktitle = {Proceedings of the SIAM International Conference on Data Mining}, address = {Lake Buena Vista, FL}, year = {2004}, pages = {333--344}, url = {http://www.cs.utexas.edu/users/ml/papers/semi-sdm-04.pdf}, keywords = {active, clustering, semi, supervised}, abstract = {Semi-supervised clustering uses a small amount of supervised data to aid unsupervised learning. One typical approach specifies a limited number of must-link and cannotlink constraints between pairs of examples. This paper presents a pairwise constrained clustering framework and a new method for actively selecting informative pairwise constraints to get improved clustering performance. The clustering and active learning methods are both easily scalable to large datasets, and can handle very high dimensional data. Experimental and theoretical results confirm that this active querying of pairwise constraints significantly improves the accuracy of clustering when given a relatively small amount of supervision.} }

%0 = article %A = Basu, Sugato and Banerjee, Arindam and Mooney, Raymond J. %B = Proceedings of the SIAM International Conference on Data Mining %C = Lake Buena Vista, FL %D = 2004 %T = Active Semi-Supervision for Pairwise Constrained Clustering %U = http://www.cs.utexas.edu/users/ml/papers/semi-sdm-04.pdf