Improving variational methods via pairwise linear response identities. Raymond, J. & Ricci-Tersenghi, F. Journal of Machine Learning Research, 18:1-36, Microtome Publishing, 2017. cited By 2
Paper abstract bibtex Inference methods are often formulated as variational approximations: these approximations allow easy evaluation of statistics by marginalization or linear response, but these estimates can be inconsistent. We show that by introducing constraints on covariance, one can ensure consistency of linear response with the variational parameters, and in so doing inference of marginal probability distributions is improved. For the Bethe approximation and its generalizations, improvements are achieved with simple choices of the constraints. The approximations are presented as variational frameworks; iterative procedures related to message passing are provided for finding the minima. ©2017 Jack Raymond and Federico Ricci-Tersenghi.
@ARTICLE{Raymond20171,
author={Raymond, J. and Ricci-Tersenghi, F.},
title={Improving variational methods via pairwise linear response identities},
journal={Journal of Machine Learning Research},
year={2017},
volume={18},
pages={1-36},
note={cited By 2},
url={https://www.scopus.com/inward/record.uri?eid=2-s2.0-85014535061&partnerID=40&md5=d2736df53c790b11c83e38800918c831},
abstract={Inference methods are often formulated as variational approximations: these approximations allow easy evaluation of statistics by marginalization or linear response, but these estimates can be inconsistent. We show that by introducing constraints on covariance, one can ensure consistency of linear response with the variational parameters, and in so doing inference of marginal probability distributions is improved. For the Bethe approximation and its generalizations, improvements are achieved with simple choices of the constraints. The approximations are presented as variational frameworks; iterative procedures related to message passing are provided for finding the minima. ©2017 Jack Raymond and Federico Ricci-Tersenghi.},
publisher={Microtome Publishing},
issn={15324435},
document_type={Article},
source={Scopus},
}
Downloads: 0
{"_id":"DdT5YGr4MMN9ZSF6v","bibbaseid":"raymond-riccitersenghi-improvingvariationalmethodsviapairwiselinearresponseidentities-2017","authorIDs":[],"author_short":["Raymond, J.","Ricci-Tersenghi, F."],"bibdata":{"bibtype":"article","type":"article","author":[{"propositions":[],"lastnames":["Raymond"],"firstnames":["J."],"suffixes":[]},{"propositions":[],"lastnames":["Ricci-Tersenghi"],"firstnames":["F."],"suffixes":[]}],"title":"Improving variational methods via pairwise linear response identities","journal":"Journal of Machine Learning Research","year":"2017","volume":"18","pages":"1-36","note":"cited By 2","url":"https://www.scopus.com/inward/record.uri?eid=2-s2.0-85014535061&partnerID=40&md5=d2736df53c790b11c83e38800918c831","abstract":"Inference methods are often formulated as variational approximations: these approximations allow easy evaluation of statistics by marginalization or linear response, but these estimates can be inconsistent. We show that by introducing constraints on covariance, one can ensure consistency of linear response with the variational parameters, and in so doing inference of marginal probability distributions is improved. For the Bethe approximation and its generalizations, improvements are achieved with simple choices of the constraints. The approximations are presented as variational frameworks; iterative procedures related to message passing are provided for finding the minima. ©2017 Jack Raymond and Federico Ricci-Tersenghi.","publisher":"Microtome Publishing","issn":"15324435","document_type":"Article","source":"Scopus","bibtex":"@ARTICLE{Raymond20171,\nauthor={Raymond, J. and Ricci-Tersenghi, F.},\ntitle={Improving variational methods via pairwise linear response identities},\njournal={Journal of Machine Learning Research},\nyear={2017},\nvolume={18},\npages={1-36},\nnote={cited By 2},\nurl={https://www.scopus.com/inward/record.uri?eid=2-s2.0-85014535061&partnerID=40&md5=d2736df53c790b11c83e38800918c831},\nabstract={Inference methods are often formulated as variational approximations: these approximations allow easy evaluation of statistics by marginalization or linear response, but these estimates can be inconsistent. We show that by introducing constraints on covariance, one can ensure consistency of linear response with the variational parameters, and in so doing inference of marginal probability distributions is improved. For the Bethe approximation and its generalizations, improvements are achieved with simple choices of the constraints. The approximations are presented as variational frameworks; iterative procedures related to message passing are provided for finding the minima. ©2017 Jack Raymond and Federico Ricci-Tersenghi.},\npublisher={Microtome Publishing},\nissn={15324435},\ndocument_type={Article},\nsource={Scopus},\n}\n\n","author_short":["Raymond, J.","Ricci-Tersenghi, F."],"key":"Raymond20171","id":"Raymond20171","bibbaseid":"raymond-riccitersenghi-improvingvariationalmethodsviapairwiselinearresponseidentities-2017","role":"author","urls":{"Paper":"https://www.scopus.com/inward/record.uri?eid=2-s2.0-85014535061&partnerID=40&md5=d2736df53c790b11c83e38800918c831"},"metadata":{"authorlinks":{}},"downloads":0,"html":""},"bibtype":"article","biburl":"http://nanotec.cnr.it/data/nanotec/scopus-2017.bib","creationDate":"2020-04-27T18:01:51.311Z","downloads":0,"keywords":[],"search_terms":["improving","variational","methods","via","pairwise","linear","response","identities","raymond","ricci-tersenghi"],"title":"Improving variational methods via pairwise linear response identities","year":2017,"dataSources":["TgRzGSnFk8ZF7cwa2","ETQm4KuYrrdfxXLmt","Xwr3Ky8YCPtt2S7S2"]}