Disentangling Gaussians.

The article discusses the Gaussian mixture model (GMM), a statistical moodel comprised of heterogeneous Gaussian sources, and presents an algorithm which is able to recover the parameters of Gaussians which has polynomial sample complexity and computational complexity. The article discusses one-dime...

Descripción completa

Detalles Bibliográficos
Publicado en:Communications of the ACM Vol. 55; no. 2; pp. 113 - 121
Autores principales: Kalai, Adam Tauman, Moitra, Ankur, Valiant, Gregory
Formato: Artículo
Publicado: Association for Computing Machinery Feb2012
Materias:
Acceso en línea:Ver este registro en EBSCOhost
Descripción
Sumario:The article discusses the Gaussian mixture model (GMM), a statistical moodel comprised of heterogeneous Gaussian sources, and presents an algorithm which is able to recover the parameters of Gaussians which has polynomial sample complexity and computational complexity. The article discusses one-dimensional GMM and discusses issues related to clustering, overlapping Gaussians, and decomposition. The authors argue that their algorithm can provide a basis for developing estimators or other algorithms with practical utilities.