Max-imum Likelihood from Incomplete Data
By Arthur P. Dempster, Nan M. Laird, Donald B. Rubin. 1972
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Yoshua Bengio
- Ruslan Salakhutdinov
- David M. Blei
- Sergey Levine
- Dacheng Tao
- Trevor Darrell
- Francis R. Bach
- Zhe Gan
Citations in the corpus
- Spectral Regularization Algorithms for Learning Large Incomplete Matrices (2010, methodology): “2001)—an EM-type (Dempster et al., 1977) iterative algorithm that alternate s between imputing the missing values from a current SVD, and updating the SVD using the “complete ” da a matrix. In its very motivation, SOFT-IMPUTE is different from generic first order algorithms (Cai et al., 2008; Ma et ”
- Functional linear discriminant analysis for irregularly sampled curves (2001, background): “However, if the ij had been observed the joint likelihood of Yij and ij would simplify to Maximizing this likelihood is much less complex which suggests treating the ij as missing data and implementing the EM algorithm (Dempster et al. (1977), Laird and Ware (1982)).”
- Generative models for discovering sparse distributed representations. (1997, background): “…di = hp(s j = 1jd)i (4) 2 i (new) = * X j p(s j = 1jd) (d i g ji ) 2 + (5) j (new) = hp(s j = 1jd)i (6) This is a version of the \Expectation and Maximization" algorithm (Dempster et al., 1977) and is guaranteed to raise the likelihood of the observed data unless it is already at a local optimum.”
- Stochastic variational inference (2012, methodology): “Our algorithm builds on the earlier approach of Sato (2001), whose algorithm only applie s to the limited set of models that can be fit with the EM algorithm (Dempster et al., 1977). Specifically , we generalize his approach to the much wider set of probabilistic models that are amenable to c losed-”
- A Note on How to Conduct a Factor-Based PLS-SEM Analysis (2015, methodology): “As a group they are akin to a “soft” version of the classic expectation-maximization algorithm (Dempster et al., 1977) used in maximum likelihood estimation, but with apparently faster convergence and nonparametric properties.”