Simple and Globally Convergent Methods for Accelerating the Convergence of Any EM Algorithm

By R. Varadhan, Christophe Roland. 2008

369 citations recorded by the source corpus.

Subfields: Control and Systems Engineering, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/97418d365f941d4c89ce33f6757b538b73cf992d

How researchers see this paper

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

In Control and Systems Engineering

In Statistics and Probability

Citations in the corpus

Majorization-Minimization Algorithms in Signal Processing, Communications, and Machine Learning (2017, methodology): “The STEM accelerator proposed in [86] approximates ∇F (x̃t) by a scaled identity matrix.”
Intelligent Reflecting Surface Aided Multigroup Multicast MISO Communication Systems (2019, methodology): “Here, we adopt SQUAREM [32] to accelerate the convergence speed of our proposed algorithm, which is summarized in Algorithm 2.”
Best practices for differentiated products demand estimation with PyBLP (2020, methodology): “33 Reynaerts et al. (2012) conduct extensive testing of various fixed point acceleration methods and find that the SQUAREM algorithm of Varadhan and Roland (2008) works well on the BLP 30A typical entry is ∂sjt ∂δkt = ∫ [1(j = k)sijt(µit) − sijt(µit)sikt(µit)]f(µit|θ̃2) dµit.”
Optimization Methods for Designing Sequences With Low Autocorrelation Sidelobes (2014, methodology): “It is the so called squared iterative method (SQUAREM), which was originally proposed in [25] to accelerate any EM algorithms.”
Sailfish enables alignment-free isoform quantification from RNA-seq reads using lightweight algorithms (2013, methodology): “Updates the relative abundance estimates according to an accelerated EM procedure whose update direction and magnitude are dynamically computed [16].”
fastSTRUCTURE: Variational Inference of Population Structure in Large SNP Data Sets (2014, methodology): “…algorithm with improved convergence rates can be derived by adapting the Cauchy–Barzilai– Borwein method for accelerating the convergence of linear fixed-point problems (Raydan and Svaiter 2002) to the nonlinear fixed-point problem given by our deterministic map (Varadhan and Roland 2008).”
Generative-Discriminative Basis Learning for Medical Imaging (2012, methodology): “There are several choices for BB step length [44], in this paper, we choose the following method to compute it [45]”
Estimating Individual Admixture Proportions from Next Generation Sequencing Data (2013, methodology): “This is the principle of squared iterative methods (Varadhan and Roland 2008) for accelerating EM algorithms.”
Fast model-based estimation of ancestry in unrelated individuals. (2009, methodology): “We also experimented with the squared extrapolation techniques of Varadhan and Roland (2008).”