Stochastic expectation maximization with variance reduction

Expectation-Maximization (EM) is a popular tool for learning latent variable models, but the vanilla batch EM does not scale to large data sets because the whole data set is needed at every E-step. Stochastic Expectation Maximization (sEM) reduces the cost of E-step by stochastic approximation. Howe...

Celý popis

Podrobná bibliografie
Hlavní autoři: Chen, J, Zhu, J, Teh, Y, Zhang, T
Médium: Conference item
Vydáno: Massachusetts Institute of Technology Press 2018