Show simple item record

dc.contributor.authorEftekhari, Armin
dc.contributor.authorHauser, Raphael A.
dc.contributor.authorGrammenos, Andreas
dc.date.accessioned2018-06-26T09:30:06Z
dc.date.available2018-06-26T09:30:06Z
dc.date.issued2018-06-26
dc.identifier.urihttp://publications.mfo.de/handle/mfo/1369
dc.description.abstractThis paper introduces Memory-limited Online Subspace Estimation Scheme (MOSES) for both estimating the principal components of data and reducing its dimension. More specifically, consider a scenario where the data vectors are presented sequentially to a user who has limited storage and processing time available, for example in the context of sensor networks. In this scenario, MOSES maintains an estimate of leading principal components of the data that has arrived so far and also reduces its dimension. In terms of its origins, MOSES slightly generalises the popular incremental Singular Value Decomposition (SVD) to handle thin blocks of data. This simple generalisation is in part what allows us to complement MOSES with a comprehensive statistical analysis that is not available for incremental SVD, despite its empirical success. This generalisationalso enables us to concretely interpret MOSES as an approximate solver for the underlying non-convex optimisation program. We also find that MOSES shows state-of-the-art performance in our numerical experiments with both synthetic and real-world datasets.en_US
dc.language.isoen_USen_US
dc.publisherMathematisches Forschungsinstitut Oberwolfachen_US
dc.relation.ispartofseriesOberwolfach Preprints;2018,12
dc.subjectPrincipal component analysisen_US
dc.subjectLinear dimensionality reductionen_US
dc.subjectSubspace identificationen_US
dc.subjectStreaming algorithmsen_US
dc.subjectNon-convex optimisationen_US
dc.titleMOSES: A Streaming Algorithm for Linear Dimensionality Reductionen_US
dc.typePreprinten_US
dc.identifier.doi10.14760/OWP-2018-12
local.scientificprogramOWLF 2017en_US
local.series.idOWP-2018-12en_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record