Learning mixtures of Gaussians
Author:
- 20 January 2003
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 02725428,p. 634-644
Abstract
Mixtures of Gaussians are among the most fundamental and widely used statistical models. Current techniques for learning such mixtures from data are local search heuristics with weak performance guarantees. We present the first provably correct algorithm for learning a mixture of Gaussians. This algorithm is very simple and returns the true centers of the Gaussians to within the precision specified by the user with high probability. It runs in time only linear in the dimension of the data and polynomial in the number of Gaussians.Keywords
This publication has 12 references indexed in Scilit:
- Estimating a mixture of two product distributions1999
- Embedding tree metrics into low dimensional Euclidean spaces1999
- Combinatorial Geometry1995
- On the learnability of discrete distributions1994
- The Johnson-Lindenstrauss lemma and the sphericity of some graphsJournal of Combinatorial Theory, Series B, 1988
- Matrix Analysis1985
- Asymptotics of Graphical Projection PursuitThe Annals of Statistics, 1984
- Mixture Densities, Maximum Likelihood and the EM AlgorithmSIAM Review, 1984
- Extensions of Lipschitz mappings into a Hilbert space1984
- Balls in Rk do not cut all subsets of k + 2 pointsAdvances in Mathematics, 1979