Skip to main content

Provable ICA with Unknown Gaussian Noise, and Implications for Gaussian Mixtures and Autoencoders

Author(s): Arora, Sanjeev; Ge, Rong; Moitra, Ankur; Sachdeva, Sushant

Download
To refer to this page use: http://arks.princeton.edu/ark:/88435/pr1885k
Full metadata record
DC FieldValueLanguage
dc.contributor.authorArora, Sanjeev-
dc.contributor.authorGe, Rong-
dc.contributor.authorMoitra, Ankur-
dc.contributor.authorSachdeva, Sushant-
dc.date.accessioned2021-10-08T19:49:52Z-
dc.date.available2021-10-08T19:49:52Z-
dc.date.issued2015en_US
dc.identifier.citationArora, Sanjeev, Rong Ge, Ankur Moitra, and Sushant Sachdeva. "Provable ICA with Unknown Gaussian Noise, and Implications for Gaussian Mixtures and Autoencoders." Algorithmica 72, no. 1 (2015): 215-236. doi:10.1007/s00453-015-9972-2en_US
dc.identifier.issn0178-4617-
dc.identifier.urihttps://dspace.mit.edu/bitstream/handle/1721.1/106898/453_2015_9972_ReferencePDF.pdf?sequence=2&isAllowed=y-
dc.identifier.urihttp://arks.princeton.edu/ark:/88435/pr1885k-
dc.description.abstractWe present a new algorithm for independent component analysis which has provable performance guarantees. In particular, suppose we are given samples of the form 𝑦=𝐴π‘₯+πœ‚ where 𝐴 is an unknown but non-singular 𝑛×𝑛 matrix, π‘₯ is a random variable whose coordinates are independent and have a fourth order moment strictly less than that of a standard Gaussian random variable and πœ‚ is an 𝑛-dimensional Gaussian random variable with unknown covariance 𝛴: We give an algorithm that provably recovers 𝐴 and 𝛴 up to an additive πœ– and whose running time and sample complexity are polynomial in 𝑛 and 1/πœ–. To accomplish this, we introduce a novel β€œquasi-whitening” step that may be useful in other applications where there is additive Gaussian noise whose covariance is unknown. We also give a general framework for finding all local optima of a function (given an oracle for approximately finding just one) and this is a crucial step in our algorithm, one that has been overlooked in previous attempts, and allows us to control the accumulation of error when we find the columns of 𝐴 one by one via local search.en_US
dc.format.extent215 - 236en_US
dc.language.isoen_USen_US
dc.relation.ispartofAlgorithmicaen_US
dc.rightsAuthor's manuscripten_US
dc.titleProvable ICA with Unknown Gaussian Noise, and Implications for Gaussian Mixtures and Autoencodersen_US
dc.typeJournal Articleen_US
dc.identifier.doi10.1007/s00453-015-9972-2-
dc.identifier.eissn1432-0541-
pu.type.symplectichttp://www.symplectic.co.uk/publications/atom-terms/1.0/journal-articleen_US

Files in This Item:
File Description SizeFormat 
ProvableICA.pdf417.26 kBAdobe PDFView/Download


Items in OAR@Princeton are protected by copyright, with all rights reserved, unless otherwise indicated.