Show simple item record

dc.contributor.authorLitvinenko, Alexander
dc.date.accessioned2017-12-11T06:28:27Z
dc.date.available2017-12-11T06:28:27Z
dc.date.issued2017-12-10
dc.identifier.urihttp://hdl.handle.net/10754/626345
dc.description.abstractMatrices began in the 2nd century BC with the Chinese. One can find traces, which go to the 4th century BC to the Babylonians. The text ``Nine Chapters of the Mathematical Art'' written during the Han Dynasty in China gave the first known example of matrix methods. They were used to solve simultaneous linear equations (more in http://math.nie.edu.sg/bwjyeo/it/MathsOnline_AM/livemath/the/IT3AMMatricesHistory.html). The first ideas of the maximum likelihood estimation (MLE) was introduces by Laplace (1749-1827), by Gauss (1777-1855), the Likelihood was defined by Thiele Thorvald (1838-1910). Why we still use matrices? The matrix data format is more than 2200 years old. Our world is multi-dimensional! Why not to introduce a more appropriate data format and why not to reformulate the MLE method for it? In this work we are utilizing the low-rank tensor formats for multi-dimansional functions, which appear in spatial statistics.
dc.description.sponsorshipKAUST
dc.subjectMLE
dc.subjectlog-likelihood
dc.subjectTucker tensor
dc.subjectlow-rank tensor approximation
dc.subjectFourier transform
dc.subjectLaplace transform
dc.subjectsinc quadrature
dc.subjectKriging in low-rank format
dc.titleStructures and algorithms for post-processing large data sets and multi-variate functions in spatio-temporal statistics
dc.typePresentation
dc.contributor.departmentExtreme Computing Research Center
dc.conference.date10.12.2017
dc.conference.nameECRC internal talk
dc.conference.locationKAUST
refterms.dateFOA2018-06-13T10:44:36Z


Files in this item

Thumbnail
Name:
litvinenko_ECRC2017.pdf
Size:
1.017Mb
Format:
PDF
Description:
Talk for the work, described in https://arxiv.org/abs/1711.06874

This item appears in the following Collection(s)

Show simple item record