MM algorithms for generalized Bradley-Terry models

By David R. Hunter. 2003

656 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/1a569c3349e6538947c4ddb0eed0307bfd5dad15

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

Citations in the corpus

Re-evaluating evaluation (2018, methodology): “Bradley-Terry models provide a general framework for pairwise comparison [39].”
A mixture of experts model for rank data with applications in election studies (2008, methodology): “Good summaries of the methodology are provided in Lange, Hunter and Yang (2000), Hunter and Lange (2004) and Hunter (2004).”
GROUPED SPARSE PAIRED COMPARISONS IN THE BRADLEY-TERRY MODEL (2011, background): “For a wide class of generalizations, Hunter (2004) proposed the iterative minorization-maximization (MM) algorithms for maximum likelihood estimation and established their good convergence properties.”
Maximum Likelihood Estimation in Log-Linear Models: Theory and Algorithms (2011, background): “roportional ﬁtting or scaling (IPS), are popular, e.g., see Fienberg (1970), Bishop et al. (1975), Darroch and Ratcliff (1972), Csiszar (1975, 1989), Lauritzen (1996) and Ruschendorf (1995). See also Hunter (2004) for the variation known as MM algorithms. Typically, these alternative algorithms are ”
Matrix estimation by Universal Singular Value Thresholding (2012, methodology): “For the basic Bradley–Terry model, it is possible to find the maximum likelihood estimate of the ai ’s using a simple iterative procedure [59, 66, 97].”
A Tutorial on MM Algorithms (2004, background): “Recent work has demonstrated the utility of MM algorithms in a broad range of statistical contexts, including quantile regression (Hunter and Lange, 2000b), survival analysis (Hunter and Lange, 2002), paired and multiple comparisons (Hunter, 2004), variable selection (Hunter and Li, 2002), and DNA s”
Less than a Single Pass: Stochastically Controlled Stochastic Gradient (2016, background): “[15] David R Hunter.”
Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference (2024, background): “Related topics include probability models (Hunter, 2004; Rao & Kupper, 1967), rank elicitation (Szörényi et al., 2015; Busa-Fekete et al., 2014a;b), and online experiment design (Chernoff, 1992; Karimi et al., 2021).”
Estimating brain functional connectivity with sparse multivariate autoregression (2005, methodology): “The speciﬁc implementation of penalized regression used in this article is that of the maximization– minorization (MM) algorithm (Hunter 2004; Hunter & Lange 2004), which exploits an optimization technique that extends the central idea of EM algorithms to situations not necessarily involving…”
Bayesian inference for Plackett-Luce ranking models (2009, methodology): “Luce can be achieved via maximum likelihood estimation (MLE) using MM methods (Hunter, 2004), we are unaware of an efficient Bayesian treatment.”
Fast and Accurate Inference of Plackett-Luce Models (2015, methodology): “As a comparison, one iteration of the MM algorithm [18] is O(D).”
Optimal full ranking from pairwise comparisons (2021, methodology): “The computation of the local MLE (24) is implemented by the MM algorithm [24].”
Spectral Method and Regularized MLE Are Both Optimal for Top-$K$ Ranking (2017, positive): “For example, Hunter (2004) examined variants of the parametric BTL model, and established the convergence properties of the minorizationmaximization algorithm for solving the corresponding MLE problem.”
Computational and Statistical Tradeoffs in Learning to Rank (2016, methodology): “ormulated as a convex optimization that can be solved eciently using the o-the-shelf solvers. However, tailored algorithms for nding the optimal solution have been pro5 posed in Ford Jr. (1957) and Hunter (2004), which iteratively nds the xed point of the KKT condition. Negahban et al. (2014) introd”