Adaptive Forward-Backward Greedy Algorithm for Sparse Learning with Linear Models

By Tong Zhang. 2008

176 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Data-driven discovery of partial differential equations (2016, methodology): “Greedy algorithms were shown to exhibit good performance on sparse optimization problems including PDE-FIND but, in some cases, were less reliable than STRidge (20).”
Stability selection (2008, methodology): “3 Randomised orthogonal Matching Pursuit An interesting alternative to Lasso or greedy forward search in this context are the recently proposed forward-backward search FOBA (Zhang, 2008) and the MC+ algorithm (Zhang, 2007), which both provably lead to consistent variable selection under weak conditi”
Multi-Label Prediction via Compressed Sensing (2009, methodology): “We tested the greedy and iterative reconstruction algorithms described earlier (OMP, FoBa, and CoSaMP) as well as a path-following version of Lasso based on LARS [21].”
Multi-Stage Multi-Task Feature Learning (2012, neutral): “We first present a sub-Gaussian noise assumption which is very common in the analysis of sparse learning literature (Zhang and Zhang, 2012; Zhang, 2008, 2009, 2010, 2012).”
Feature Selection as a One-Player Game (2010, methodology): “Hill-climbing approaches (Hall, 2000), mixing forward and backward selection (Zhang, 2008a), using look-ahead or making large steps in the lattice (Margaritis, 2009), combining global and local search operators (Boull´e, 2007), among many other heuristics, have been proposed to address the EvE…”
A Regularization Approach to Nonlinear Variable Selection (2010, background): “In particular Zhang (2009) applies different subset selection techniques and compares their prediction accuracy.”
Submodular meets Spectral: Greedy Algorithms for Subset Selection, Sparse Approximation and Dictionary Selection (2011, methodology): “The goal is to select a dictionary D of d observation variables, to optimize the average R 2 ﬁt for the Z i using at most k vectors from D for each.”
Features in Concert: Discriminative Feature Selection meets Unsupervised Clustering (2014, methodology): “We evaluated six methods: ours, SVM on all input features, AdaBoost on all input features, ours with SVM (applying SVM only to features selected by our method, idea related to [22, 37, 47]), forward-backward selection (FoBa) [50] and simple averaging over all input features.”
Sparse Quantile Huber Regression for Efficient and Robust Estimation (2014, methodology): “More recently, greedy selection methods such as Orthogonal Matching Pursuit (OMP) [20] have received considerable attention [37] [12] as an alternative to the ℓ1-penalized methods.”
Trading Accuracy for Sparsity in Optimization Problems with Sparsity Constraints (2010, methodology): “One of the strongest results is the one derived for the FoBa algorithm of [33].”
Restricted Strong Convexity Implies Weak Submodularity (2016, methodology): “Forward Backward (FoBa) [52] first runs FS at each step and then drops any features if doing so would decrease the objective by less than half the latest marginal gain.”