loading

Logout succeed

Logout succeed. See you again!

ebook img

An Almost Optimal Algorithm for Computing Nonnegative Rank PDF

pages68 Pages
release year2013
file size0.33 MB
languageEnglish

Preview An Almost Optimal Algorithm for Computing Nonnegative Rank

An Almost Optimal Algorithm for Computing Nonnegative Rank Ankur Moitra Institute for Advanced Study January 8, 2013 AnkurMoitra (IAS,PU) Provably January8,2013 m M n W m M = A inner−dimension n rank W m M = A inner−dimension n rank non−negative W m M = A inner−dimension n non−negative n o n − n e g a t i vraenk non−negative W m M = A inner−dimension n non−negative Combinatorics: extended formulation, log-rank conjecture [Yannakakis], [Lov´asz, Saks] Physical Modeling: interaction of components is additive visual recognition, environmetrics Applications Statistics and Machine Learning: extract latent relationships in data image segmentation, text classification, information retrieval, collaborative filtering, ... [Lee, Seung], [Xu et al], [Hofmann], [Kumar et al], [Kleinberg, Sandler] Physical Modeling: interaction of components is additive visual recognition, environmetrics Applications Statistics and Machine Learning: extract latent relationships in data image segmentation, text classification, information retrieval, collaborative filtering, ... [Lee, Seung], [Xu et al], [Hofmann], [Kumar et al], [Kleinberg, Sandler] Combinatorics: extended formulation, log-rank conjecture [Yannakakis], [Lov´asz, Saks] Applications Statistics and Machine Learning: extract latent relationships in data image segmentation, text classification, information retrieval, collaborative filtering, ... [Lee, Seung], [Xu et al], [Hofmann], [Kumar et al], [Kleinberg, Sandler] Combinatorics: extended formulation, log-rank conjecture [Yannakakis], [Lov´asz, Saks] Physical Modeling: interaction of components is additive visual recognition, environmetrics [Vavasis]: It is NP-hard to compute the nonnegative rank. [Cohen and Rothblum]: The nonnegative rank can be computed in time (nm)O(nr+mr). [Arora, Ge, Kannan and Moitra]: The nonnegative rank can be computed in time (nm)f(r) where f(r)=O(2r) and any algorithm that runs in time (nm)o(r) would yield a sub exponential time algorithm for 3-SAT. Theorem The nonnegative rank can be computed in time (nm)O(r2). ...these algorithms are about an algebraic question, about how to best encode nonnegative rank as a systems of polynomial inequalities The Complexity of Nonnegative Rank

See more

The list of books you might like