Linked Data Explorerhttp://worldcat.org/entity/work/id/2184922570
Learning probability distributions
The first part of this thesis presents an algorithm which takes data from an unknown mixture of Gaussians in arbitrarily high dimension and recovers the parameters of this mixture to within the precision desired by the user. There are two restrictions on the mixture: its component Gaussians must be "well-separated" in a precise sense, and they must have a common (though of course unknown) non-singular covariance matrix. The running time of the algorithm is linear in the dimension of the data and polynomial in the number of Gaussians.
Open All Close All
http://schema.org/description
- "The first part of this thesis presents an algorithm which takes data from an unknown mixture of Gaussians in arbitrarily high dimension and recovers the parameters of this mixture to within the precision desired by the user. There are two restrictions on the mixture: its component Gaussians must be "well-separated" in a precise sense, and they must have a common (though of course unknown) non-singular covariance matrix. The running time of the algorithm is linear in the dimension of the data and polynomial in the number of Gaussians."@en
http://schema.org/name
- "Learning probability distributions"@en