A Monte Carlo Implementation of the EM Algorithm and the Poor Man's Data Augmentation Algorithms

By Greg C. G. Wei, Martin A. Tanner. 1990

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Fast Gradient-Based Inference with Continuous Latent Variable Models in Auxiliary Form (2013, methodology): “atent variables, like Z 1 !Z 2 !Z 3 !X. The parameters follow a zero-centered Gaussian prior distribution: N(:;0;0:01I), except for the ˙’s which follow logN(:;0;0:01I). We used Monte Carlo EM (MCEM) [WT90] for learning the parameters of this generative network, where Hybrid Monte Carlo [DKPR87] was”
Inferring High-Level Behavior from Low-Level Sensors (2003, methodology): “Our approach is in fact a direct extension of the Monte Carlo EM algorithm [12].”
Missing data methods in longitudinal studies: a review (2009, methodology): “The Monte Carlo EM algorithm given by Wei and Tanner (1990) requires generating a sample from [ymis,i,bi|yobs,i,ri,γ (t)] for each i.”
ZhuSuan: A Library for Bayesian Deep Learning (2017, methodology): “For parameter estimation of LVMs, the samples can be used in the Monte-Carlo EM (MCEM) algorithm (Wei and Tanner, 1990).”
Estimation of Affine Term Structure Models with Spanned or Unspanned Stochastic Volatility (2014, methodology): “We use the Monte Carlo Expectation Maximization (MCEM) algorithm to estimate the model, see Wei and Tanner(1990).”
Are Product Returns a Necessary Evil? Antecedents and Consequences (2009, methodology): “To do this, we augment the censored data by drawing values from a truncated normal distribution for the model estimation (Wei and Tanner 1990).”
Likelihood-Based Methods for Missing Covariates in the Cox Proportional Hazards Model (2001, methodology): “However, note that (13) is an expectation with respect to 6zmis1i — zobs1i1xi1„i1ˆ 4m57 and may be evaluated using the Monte Carlo EM of Wei and Tanner (1990) and Ibrahim, Chen, and Lipsitz (1999).”
Randomized Optimum Models for Structured Prediction (2012, methodology): “Monte Carlo expectation maximization (MCEM) (Wei and Tanner, 1990) alternates between computing a Monte Carlo estimate of the expected complete data log likelihood (the E-step) and maximizing ψ given this approximation (the M-step).”
Asymmetric heavy-tailed vector auto-regressive processes with application to financial data (2020, methodology): “In this case, the corresponding algorithm is called the MCECME algorithm; see Wei and Tanner [27], McLachlan and Krishnan [28] and Zeller et al. [25].”
Efficient indexing for large scale visual search (2009, methodology): “Because the maximum likelihood estimation of IDM is intractable [3, 5], we adopt the Monte Carlo EM algorithm to estimate parameters [7].”
Particle gibbs with ancestor sampling (2014, methodology): “When the above integral is intractable to compute, one can use a Monte Carlo approximation or a stochastic approximation of the intermediate quantity, leading to the MCEM (Wei and Tanner, 1990) and the SAEM (Delyon et al.”
Monte Carlo State-Space Likelihoods by Weighted Posterior Kernel Density Estimation (2004, methodology): “Two of the main approaches in the literature are MCEM (Wei and Tanner 1990; Chan and Ledolter 1995; McCulloch 1997; Robert and Casella 1999; Booth and Hobert 1999) and MCLR (Geyer and Thompson 1992, Geyer 1994, 1996; McCulloch 1997), which both use alternating steps of MC sampling and local…”
Hierarchical kernel stick-breaking process for multi-task image analysis (2008, methodology): “Alternatively, motivated by the Monte Carlo Expectation Maximization (MCEM) al-gorithm (Wei & Tanner, 1990), we develop a Monte Carlo Variational Bayesian (MCVB) inference algorithm, where the intractable nodes are approximated with Monte Carlo samples from their conditional posterior distributions.”