Coupling a stochastic approximation version of EM with an MCMC procedure

By Estelle Kuhn, Marc Lavielle. 2004

316 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

How researchers see this paper

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

In Artificial Intelligence

In Statistics and Probability

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.”
Stochastic Approximation Methods for Latent Regression Item Response Models (2010, background): “EM-algorithm. Gu and Zhu (2001) as well as Kuhn and Lavielle (2004) showed how”
Maximum Likelihood Estimation for Stochastic Differential Equations with Random Effects (2013, methodology): “Alternative methods have also been developed such as the SAEM algorithm (Kuhn and Lavielle, 2004).”
Sparse stochastic inference for latent Dirichlet allocation (2012, methodology): “Kuhn & Lavielle (2004) extend SAEM to use MCMC estimates.”
An efficient stochastic approximation EM algorithm using conditional particle filters (2013, background): “Instead, it is sufficient to sample from a family of Markov kernels {M θ : θ ∈ Θ}, leaving the family of posteriors invariant [19].”
Adaptive sequential Monte Carlo by means of mixture of experts (2011, methodology): “A convergence proof of this stochastic approximation algorithm will be the topic of the aforementioned companion paper and is rooted in the analysis of generic Online EM (Cappé and Moulines, 2009), stochastic approximation EM (SAEM, Delyon et al., 1999; Kuhn and Lavielle, 2004; Andrieu et al., 2005)”
Identification of jump Markov linear models using particle filters (2014, methodology): “In particular, under certain smoothness and ergodicity conditions, the sequence of iterates {θk}k≥1 will converge to a maximizer of pθ(y1:T ) as k → ∞, regardless of the number of particles N ≥ 2 used in the internal CPF-AS procedure (see [6, Proposition 1] together with [27] for details).”