Maximally Informative Statistics Export

@misc{citeulike:410963, abstract = {In this paper we propose a Bayesian, information theoretic approach to dimensionality reduction. The approach is formulated as a variational principle on mutual information, and seamlessly addresses the notions of sufficiency, relevance, and representation. Maximally informative statistics are shown to minimize a Kullback-Leibler distance between posterior distributions. Illustrating the approach, we derive the maximally informative one dimensional statistic for a random sample from the Cauchy distribution.}, archivePrefix = {arXiv}, author = {Wolf, David R. and George, Edward I.}, citeulike-article-id = {410963}, citeulike-linkout-0 = {http://arxiv.org/abs/physics/0010039}, citeulike-linkout-1 = {http://arxiv.org/pdf/physics/0010039}, comment = {finds statistic k(.) which minimizes KL-divergence between p(f|data) and p(f|k(data))}, day = {15}, eprint = {physics/0010039}, keywords = {bayesian, feature-selection}, month = {Oct}, posted-at = {2005-11-29 02:41:57}, priority = {2}, title = {Maximally Informative Statistics}, url = {http://arxiv.org/abs/physics/0010039}, year = {2000} }

TY - ELEC ID - citeulike:410963 L3 - citeulike-article-id:410963 N2 - In this paper we propose a Bayesian, information theoretic approach to dimensionality reduction. The approach is formulated as a variational principle on mutual information, and seamlessly addresses the notions of sufficiency, relevance, and representation. Maximally informative statistics are shown to minimize a Kullback-Leibler distance between posterior distributions. Illustrating the approach, we derive the maximally informative one dimensional statistic for a random sample from the Cauchy distribution. TI - Maximally Informative Statistics KW - bayesian KW - feature-selection N1 - finds statistic k(.) which minimizes KL-divergence between p(f|data) and p(f|k(data)) AU - Wolf, David AU - George, Edward PY - 2000/Oct/15/ UR - http://arxiv.org/abs/physics/0010039 ER -