Regression Shrinkage and Selection via the Lasso

By Robert Tibshirani. 1996

How researchers see this paper

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

In Optimization & Training

Citations in the corpus

Minding the Gaps for Block Frank-Wolfe Optimization of Structured SVMs (2016, background): “(2004), in their seminal paper, introduced the notion of regularization path and showed that the regularization path of LASSO (Tibshirani, 1996) is piecewise linear.”
Optimization Methods for Large-Scale Machine Learning (2016, methodology): “ applications. However, in our presentation of optimization methods, we assume that has been prescribed and is xed. We remark in passing that (8.2) has as a special case the well-known LASSO problem [151] when F(w) = kAw bk2 2 for some A2Rn dand b2Rn. Although nondierentiable, the regularized ‘ 1 pr”
Deep Interest Network for Click-Through Rate Prediction (2017, background): “ with addition of ne-grained user visited goodidsfeature, model performance falls rapidly after the rst epoch. Many methods have been proposed to reduce overtting, such as L 2 and L 1 regularization [15], and Dropout [16]. However, with sparse and high dimensional data, CTR prediction task faces gre”
A Comprehensive guide to Bayesian Convolutional Neural Network with Variational Inference (2019, background): “L 1 norm is convex and hence diﬀerentiable and can be used as an approximation to L 0 norm [59].”
Hyperparameter optimization with approximate gradient (2016, background): “Among its most well-known instances are the LASSO (Tibshirani, 1996), in which (cid:96) 1 regularization is added to a squared loss to encourage sparsity in the solutions, or (cid:96) 2 -regularized logistic regression, in which squared (cid:96) 2 regularization (known as weight decay in the context”
RAPPOR: Randomized Aggregatable Privacy-Preserving Ordinal Response (2014, methodology): “Use Lasso [26] regression to ﬁt a model Y ∼ X and select candidate strings corresponding to non-zero coeﬃcients.”
Is Feature Selection Secure against Training Data Poisoning? (2015, methodology): “After deriving and re-arranging in matrix form, one obtains: where The term v yields zero for LASSO, the identity matrix I for ridge, and (1 − ρ ) I for the elastic net.”
Functional Mechanism: Regression Analysis under Differential Privacy (2012, methodology): “Although regularization is mostly used in regression analysis to avoid overﬁtting [14,29], it also helps achieving a bounded ¯ f D ( ω ) .”
The mythos of model interpretability (2016, background): “Note, however, that humans exhibit none of these forms of transparency.”
Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) (2017, background): “This natural, expected property may be beneﬁcially used to make ﬁne-grained distinctions since relative comparisons between related concepts are a good interpretative tool (Kim et al., 2015; Doshi-Velez et al., 2015; Tibshirani, 1994; Salvatore et al., 2014).”
Interpretable Recurrent Neural Networks Using Sequential Sparse Recovery (2016, methodology): “Problem (16) is known as basis pursuit denoising (BPDN) [15], which is also equivalent to minimizing the Lagrangian of the least absolute shrinkage and selection operator (LASSO) method for sparse recovery [16].”
Convergent Learning: Do different neural networks learn the same representations? (2015, methodology): “We can relax this one-to-one constraint to various degrees by learning a mapping layer with an L1 penalty (known as a LASSO model, (Tibshirani, 1996)), where stronger penalties will lead to sparser (more few-to-one or one-to-one) mappings.”
The Bayesian Case Model: A Generative Approach for Case-Based Reasoning and Prototype Classification (2014, methodology): “In our view, there are at least three widely known types of interpretable models: sparse linear classifiers ([30, 8, 31]); discretization methods, such as decision trees and decision lists (e.”
Interpretable classification models for recidivism prediction (2015, methodology): “L1-penalized methods are typically used to create linear models that are sparse (Tibshirani, 1996; Hesterberg et al., 2008).”