Convergence of a stochastic approximation version of the EM algorithm

By Bernard Delyon, Marc Lavielle, Eric Moulines. 1999

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

Causal Discovery from Temporally Aggregated Time Series (2017, methodology): “To reduce the number of simulated sample points, we propose to use the stochastic approximation EM (SAEM) algorithm (Delyon et al., 1999), which only requires of a single realization of the unobserved variables at each iteration.”
Identification of jump Markov linear models using particle filters (2014, methodology): “To address this shortcoming, [24] proposed to use a stochastic approximation (SA) [25] of the intermediate quantity instead of a vanilla Monte Carlo approximation, resulting in the stochastic approximation EM (SAEM) algorithm.”
Stochastic Approximation Methods for Latent Regression Item Response Models (2010, methodology): “Gu and Kong (1998) have demonstrated the use of stochastic approximation with a Markov-chain generated by a MH sampler, as have Delyon et al. (1999) for the EM algorithm. Gu and Zhu (2001) as well as Kuhn and Lavielle (2004) have shown how to combine Markov-chain Monte-Carlo (MCMC) and stochastic ap”
Bounded Approximations for Marginal Likelihoods (2010, methodology): “Similar to application in variational MLE with missing data (Celeux and Diebolt, 1992; Delyon et al., 1999), stochastic approximation (SA) algorithms based on an iterative Monte Carlo procedures can be used in cases where the expectation step in the EM algorithm can-”
Specific and Shared Causal Relation Modeling and Mechanism-Based Clustering (2019, methodology): “In particular, we exploit a stochastic approximation expectation maximization (SAEM) algorithm [4], combined with Gibbs sampling in the E step and EM algorithm in the M step, for model estimation.”
An efficient stochastic approximation EM algorithm using conditional particle filters (2013, methodology): “To be able to make more efficient use of the simulated variables , a related method, referred to as stochastic approximation EM (SAEM) was proposed by [11].”
Causal Discovery and Forecasting in Nonstationary Environments with State-Space Models (2019, methodology): “Under appropriate assumptions, SAEM is shown to converge for fixed M , as k → ∞ (Delyon et al., 1999).”
Coupling a stochastic approximation version of EM with an MCMC procedure (2004, methodology): “In this situation, the assumptions of [5] that ensure the convergence of SAEM are no longer satisfied.”
A unifying framework for partial volume segmentation of brain MR images (2003, unknown): “algorithm [30] instead, in which simulated data from previous”
Central Limit Theorems for Stochastic Approximation with controlled Markov chain dynamics (2013, result): “The proof is on the same lines as the proof of ([12], Lem.”
Metropolis-Hastings Robbins-Monro Algorithm for Confirmatory Item Factor Analysis (2010, methodology): “In certain cases, it is equivalent to the stochastic approximation expectation-maximization (SAEM) algorithm of Delyon, Lavielle, and Moulines (1999).”
Sparse stochastic inference for latent Dirichlet allocation (2012, methodology): “Stochastic approximation EM (SAEM, Delyon et al., 1999) combines an EM algorithm with a stochastic online inference procedure.”
Fast derivatives of likelihood functionals for ODE based models using adjoint-state method (2016, methodology): “The evaluation of gradients becomes very costly and consequently various derivative-free methods have become more popular, see for example Delyon et al. (1999).”
A Correlated Probit Model for Joint Modeling of Clustered Binary and Continuous Responses (2001, methodology): “…for a complete data loglikelihood of the form log Lu4–5 D 6a4–57T z4u5 ƒ b4y1–51 where z4u5 is a vector function of the complete data u. Liao (1999) and Delyon et al. (1999) proposed replacing the usual Monte Carlo EM E step by calculating Nz4r5 D 41ƒ wr 5Nz4r ƒ15 C wr 1 mr mrX kD1 z4u4k55…”