Fast Moment-Based Estimation for Hierarchical Models

Perry, Patrick O.

doi:10.1111/rssb.12165

Cited by 37 publications

(41 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DEM was implemented in R (R Core Team 2016) using the Rmpi package (Yu 2002) and van Dyk's ECME algorithm. Our real data analysis based on the examples considered in Perry (2017) showed that DEM with γ = 0.3, 0.5, 0.7 matches the accuracy of van Dyk's ECME in parameter estimation while being significantly faster for all three values of γ; see Section 5. The major advantage of DEM was its generality in that it scaled the ECME algorithm to massive data settings using its non-distributed implementation and MPI.…”

Section: Motivating Example: Movielens Ratings Datamentioning

confidence: 74%

An Asynchronous Distributed Expectation Maximization Algorithm for Massive Data: The DEM Algorithm

Srivastava

DePalma

Liu

2018

Journal of Computational and Graphical Statistics

View full text Add to dashboard Cite

The family of Expectation-Maximization (EM) algorithms provides a general approach to fitting flexible models for large and complex data. The expectation (E) step of EM-type algorithms is time consuming in massive data applications because it requires multiple passes through the full data. We address this problem by proposing an asynchronous and distributed generalization of the EM called the Distributed EM (DEM). Using DEM, existing EM-type algorithms are easily extended to massive data settings by exploiting the divide-and-conquer technique and widely available computing power, such as grid computing. The DEM algorithm reserves two groups of computing processes called workers and managers for performing the E step and the maximization step (M step), respectively. The samples are randomly partitioned into a large number of disjoint subsets and are stored on the worker processes. The E step of DEM algorithm is performed in parallel on all the workers, and every worker communicates its results to the managers at the end of local E step. The managers perform the M step after they have received results from a γ-fraction of the workers, where γ is a fixed constant in (0, 1].The sequence of parameter estimates generated by the DEM algorithm retains the attractive properties of EM: convergence of the sequence of parameter estimates to a local mode and linear global rate of convergence. Across diverse simulations focused on linear mixed-effects models, the DEM algorithm is significantly faster than competing EM-type algorithms while having a

show abstract

Section: Motivating Example: Movielens Ratings Datamentioning

confidence: 74%

An Asynchronous Distributed Expectation Maximization Algorithm for Massive Data: The DEM Algorithm

Srivastava

DePalma

Liu

2018

Journal of Computational and Graphical Statistics

View full text Add to dashboard Cite

show abstract

“…Using the empirical marker distribution, the model generates the log-odds that the expression of a given marker is predictive of the sample type (for example, drug-treated vs. placebo-treated) with the 95% confidence intervals. For paired comparisons, we computed p-values using the asymptotic theory implemented in R package mbest (Perry 2017). For unpaired comparisons, we computed p-values by inverting the percentile bootstrap confidence intervals and assuming two-sided intervals with equal tails (Efron and Tibshirani 1993).…”

Section: Discussionmentioning

confidence: 99%

Characterization of the impact of daclizumab beta on circulating natural killer cells by mass cytometry

Ranganath

Simpson

Seiler

et al. 2019

Preprint

View full text Add to dashboard Cite

AbstractDaclizumab beta is a humanized monoclonal antibody that binds to CD25 and selectively inhibits high-affinity IL-2 receptor signaling. As a former treatment for relapsing forms of multiple sclerosis (RMS), daclizumab beta induces robust expansion of the CD56bright subpopulation of NK cells that is correlated with the drug’s therapeutic effects. As NK cells represent a heterogeneous population of lymphocytes with a range of phenotypes and functions, the goal of this study was to better understand how daclizumab beta altered the NK cell repertoire to provide further insight into the possible mechanism(s) of action in RMS. We used mass cytometry to evaluate expression patterns of NK cell markers and provide a comprehensive assessment of the NK cell repertoire in individuals with RMS treated with daclizumab beta or placebo over the course of one year. Treatment with daclizumab beta significantly altered the NK cell repertoire compared to placebo treatment. As previously reported, daclizumab beta significantly increased expression of CD56 on total NK cells. Within the CD56bright NK cells, treatment was associated with multiple phenotypic changes, including increased expression of NKG2A and NKp44, and diminished expression of CD244, CD57, and NKp46. While the changes were less dramatic, CD56dim NK cells responded distinctly to daclizumab beta treatment, with higher expression of CD2 and NKG2A, and lower expression of FAS-L, HLA-DR, NTB-A, NKp30, and Perforin. Together, these data indicate that the expanded NK cells share features of both immature and mature NK cells. These findings show that daclizumab beta treatment is associated with unique changes in NK cells that may enhance their ability to kill autoreactive T cells or to exert immunomodulatory functions.

show abstract

“…We use the semi-weighted scheme for choosing W ij as described by (Perry, 2017). Now, the moment equations have consistent dimension across all nodes: for every node ij ∈ N l , equation (10) has dimension p l−1 × 1, and (11) has dimension q l × q l .…”

Section: 2mentioning

confidence: 99%

“…Let Perry's (2017) results imply that, subject to assumptions on the sample size, conditional on b i , the quantity Ω…”

Section: 2mentioning

confidence: 99%

Fitting a deeply nested hierarchical model to a large book review dataset using a moment-based estimator

Zhang¹,

Schmaus²,

Perry³

2019

Ann. Appl. Stat.

Self Cite

View full text Add to dashboard Cite

We consider a particular instance of a common problem in recommender systems: using a database of book reviews to inform usertargeted recommendations. In our dataset, books are categorized into genres and sub-genres. To exploit this nested taxonomy, we use a hierarchical model that enables information pooling across across similar items at many levels within the genre hierarchy. The main challenge in deploying this model is computational: the data sizes are large, and fitting the model at scale using off-the-shelf maximum likelihood procedures is prohibitive. To get around this computational bottleneck, we extend a moment-based fitting procedure proposed for fitting single-level hierarchical models to the general case of arbitrarily deep hierarchies. This extension is an order of magnetite faster than standard maximum likelihood procedures. The fitting method can be deployed beyond recommender systems to general contexts with deeply-nested hierarchical generalized linear mixed models.

show abstract

Fast Moment-Based Estimation for Hierarchical Models

Cited by 37 publications

References 41 publications

An Asynchronous Distributed Expectation Maximization Algorithm for Massive Data: The DEM Algorithm

An Asynchronous Distributed Expectation Maximization Algorithm for Massive Data: The DEM Algorithm

Characterization of the impact of daclizumab beta on circulating natural killer cells by mass cytometry

Fitting a deeply nested hierarchical model to a large book review dataset using a moment-based estimator

Contact Info

Product

Resources

About