Dual coordinate descent methods for logistic regression and maximum entropy models

By Hsiang-Fu Yu, Fang-Lan Huang, Chih-Jen Lin. 2011

411 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Numerical Analysis, Statistical and Nonlinear Physics

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 Numerical Analysis

In Statistical and Nonlinear Physics

Citations in the corpus

Interpretable Predictions of Tree-based Ensembles via Actionable Feature Tweaking (2017, methodology): “In their work, Logistic Regression (LogReg) [26] was the”
Distributed Newton Methods for Regularized Logistic Regression (2015, methodology): “[19], the sub-problem (3) is approximately solved by a coordinate descent method [17] with a fixed number of iterations, where the number of iterations is decided by a validation procedure.”
Scaling Up Stochastic Dual Coordinate Ascent (2015, methodology): “• LIBLIN: LibLinear toolkit [25] implementation of SDCA1, swept by one of its authors on different hardware with comparable characteristics;”
Scalable Coordinate Descent Approaches to Parallel Matrix Factorization for Recommender Systems (2012, methodology): “Recently it has been successfully applied to various large-scale problems such as linear SVMs [14], maximum entropy models [15], NMF problems [9], [10], and sparse inverse covariance estimation”
Population-Level Prediction of Type 2 Diabetes From Claims Data and Analysis of Risk Factors (2015, positive): “36,38 We use an algorithm based on Dual Coordinate Descent, 38 which handles massive datasets very efﬁciently, to train these models from data.”
Recent Advances of Large-Scale Linear Classification (2012, methodology): “The work in [29] follows [64] to apply a two-level coordinate descent method, but uses a different method in the second level to decide variables for update.”
LIBLINEAR: A Library for Large Linear Classification (2008, methodology): “L2-regularized Logistic Regression (Solving Dual) See Yu et al. (2011) for details of a dual coordinate descent method.”
PASSCoDe: Parallel ASynchronous Stochastic dual Co-ordinate Descent (2015, background): “For SVM problems, the above subproblem has a closed form solution, while for logistic regression problems it has to be solved by an iterative solver (see (Yu et al., 2011) for details).”
Distributed coordinate descent for generalized linear models with regularization (2016, background): “Coordinate descent algorithms works well in primal [1] and dual [11].”
What makes a great MOOC? An interdisciplinary analysis of student retention in online courses (2013, methodology): “all multiclass strategy with logistic regression (Yu et al. 2011) and ridge regression.”
A Workflow for Visual Diagnostics of Binary Classifiers using Instance-Level Explanations (2017, methodology): “Training Test AUC Gaussian Na¨ıve Bayes (GNB) [33] 0.58 0.52 Logistic Regression (LR) [32] 0.85 0.79 Random Forest (RF) [6] 0.88 0.79 Multi Layer Perceptron (MLP) [14] 0.85 0.80 As we can see most of the models achieve similar performance on the test data.”
Promoting Positive Post-Click Experience for In-Stream Yahoo Gemini Users (2015, methodology): “Using the features listed above we train several models to predict P i HQ using three well-known learning methods: Logistic Regression ( LogReg ) [28], Support Vector Machines ( SVM ) [7], and Gradient Boosted Decision Trees ( GBDT ) [10].”