A statistical learning theory framework for supervised pattern discovery. Huggins, J. H & Rudin, C. *Proc. of SIAM International Conference on Data Mining (SDM)*, 2014. Paper abstract bibtex This paper formalizes a latent variable inference problem we call supervised pattern discovery, the goal of which is to find sets of observations that belong to a single ``pattern.'' We discuss two versions of the problem and prove uniform risk bounds for both. In the first version, collections of patterns can be generated in an arbitrary manner and the data consist of multiple labeled collections. In the second version, the patterns are assumed to be generated independently by identically distributed processes. These processes are allowed to take an arbitrary form, so observations within a pattern are not in general independent of each other. The bounds for the second version of the problem are stated in terms of a new complexity measure, the quasi-Rademacher complexity.

@article{HuRu-SDM-2014,
Abstract = {This paper formalizes a latent variable inference problem we call supervised pattern discovery, the goal of which is to find sets of observations that belong to a single ``pattern.'' We discuss two versions of the problem and prove uniform risk bounds for both. In the first version, collections of patterns can be generated in an arbitrary manner and the data consist of multiple labeled collections. In the second version, the patterns are assumed to be generated independently by identically distributed processes. These processes are allowed to take an arbitrary form, so observations within a pattern are not in general independent of each other. The bounds for the second version of the problem are stated in terms of a new complexity measure, the quasi-Rademacher complexity.},
Author = {Huggins, Jonathan H and Rudin, Cynthia},
Date-Added = {2014-11-28 03:53:41 +0000},
Date-Modified = {2015-05-19 13:29:24 +0000},
Journal = {Proc. of SIAM International Conference on Data Mining (SDM)},
Keywords = {learning theory},
Title = {A statistical learning theory framework for supervised pattern discovery},
Url = {http://arxiv.org/abs/1307.0802},
Year = {2014},
Bdsk-Url-1 = {../papers/HuRu-SDM-2014.pdf}}

Downloads: 0

{"_id":"QjmCGWHRnPWkZvjSn","bibbaseid":"huggins-rudin-astatisticallearningtheoryframeworkforsupervisedpatterndiscovery-2014","downloads":0,"creationDate":"2015-12-06T02:41:18.430Z","title":"A statistical learning theory framework for supervised pattern discovery","author_short":["Huggins, J. H","Rudin, C."],"year":2014,"bibtype":"article","biburl":"http://web.mit.edu/jhuggins/www/files/mypubs.bib","bibdata":{"bibtype":"article","type":"article","abstract":"This paper formalizes a latent variable inference problem we call supervised pattern discovery, the goal of which is to find sets of observations that belong to a single ``pattern.'' We discuss two versions of the problem and prove uniform risk bounds for both. In the first version, collections of patterns can be generated in an arbitrary manner and the data consist of multiple labeled collections. In the second version, the patterns are assumed to be generated independently by identically distributed processes. These processes are allowed to take an arbitrary form, so observations within a pattern are not in general independent of each other. The bounds for the second version of the problem are stated in terms of a new complexity measure, the quasi-Rademacher complexity.","author":[{"propositions":[],"lastnames":["Huggins"],"firstnames":["Jonathan","H"],"suffixes":[]},{"propositions":[],"lastnames":["Rudin"],"firstnames":["Cynthia"],"suffixes":[]}],"date-added":"2014-11-28 03:53:41 +0000","date-modified":"2015-05-19 13:29:24 +0000","journal":"Proc. of SIAM International Conference on Data Mining (SDM)","keywords":"learning theory","title":"A statistical learning theory framework for supervised pattern discovery","url":"http://arxiv.org/abs/1307.0802","year":"2014","bdsk-url-1":"../papers/HuRu-SDM-2014.pdf","bibtex":"@article{HuRu-SDM-2014,\n\tAbstract = {This paper formalizes a latent variable inference problem we call supervised pattern discovery, the goal of which is to find sets of observations that belong to a single ``pattern.'' We discuss two versions of the problem and prove uniform risk bounds for both. In the first version, collections of patterns can be generated in an arbitrary manner and the data consist of multiple labeled collections. In the second version, the patterns are assumed to be generated independently by identically distributed processes. These processes are allowed to take an arbitrary form, so observations within a pattern are not in general independent of each other. The bounds for the second version of the problem are stated in terms of a new complexity measure, the quasi-Rademacher complexity.},\n\tAuthor = {Huggins, Jonathan H and Rudin, Cynthia},\n\tDate-Added = {2014-11-28 03:53:41 +0000},\n\tDate-Modified = {2015-05-19 13:29:24 +0000},\n\tJournal = {Proc. of SIAM International Conference on Data Mining (SDM)},\n\tKeywords = {learning theory},\n\tTitle = {A statistical learning theory framework for supervised pattern discovery},\n\tUrl = {http://arxiv.org/abs/1307.0802},\n\tYear = {2014},\n\tBdsk-Url-1 = {../papers/HuRu-SDM-2014.pdf}}\n","author_short":["Huggins, J. H","Rudin, C."],"key":"HuRu-SDM-2014","id":"HuRu-SDM-2014","bibbaseid":"huggins-rudin-astatisticallearningtheoryframeworkforsupervisedpatterndiscovery-2014","role":"author","urls":{"Paper":"http://arxiv.org/abs/1307.0802"},"keyword":["learning theory"],"downloads":0,"html":"","metadata":{"authorlinks":{"huggins, j":"http://web.mit.edu/"}}},"search_terms":["statistical","learning","theory","framework","supervised","pattern","discovery","huggins","rudin"],"keywords":["learning theory"],"authorIDs":["Z3NW3a7JPsgeLxRAX"],"dataSources":["mSG4eqt2wgj6d5mz3"]}