Raftul cu initiativa Book Archive

Machine Theory

Statistical learning with sparsity : the lasso and by Hastie, Trevor

By Hastie, Trevor

Show description

Read or Download Statistical learning with sparsity : the lasso and generalizations. PDF

Similar machine theory books

Digital and Discrete Geometry: Theory and Algorithms

This publication offers finished insurance of the fashionable equipment for geometric difficulties within the computing sciences. It additionally covers concurrent subject matters in information sciences together with geometric processing, manifold studying, Google seek, cloud information, and R-tree for instant networks and BigData. the writer investigates electronic geometry and its similar optimistic tools in discrete geometry, delivering targeted tools and algorithms.

Artificial Intelligence and Symbolic Computation: 12th International Conference, AISC 2014, Seville, Spain, December 11-13, 2014. Proceedings

This e-book constitutes the refereed lawsuits of the twelfth overseas convention on man made Intelligence and Symbolic Computation, AISC 2014, held in Seville, Spain, in December 2014. The 15 complete papers offered including 2 invited papers have been conscientiously reviewed and chosen from 22 submissions.

Statistical Language and Speech Processing: Third International Conference, SLSP 2015, Budapest, Hungary, November 24-26, 2015, Proceedings

This booklet constitutes the refereed court cases of the 3rd overseas convention on Statistical Language and Speech Processing, SLSP 2015, held in Budapest, Hungary, in November 2015. The 26 complete papers offered including invited talks have been conscientiously reviewed and chosen from seventy one submissions.

Additional info for Statistical learning with sparsity : the lasso and generalizations.

Sample text

We are using the spelling in the original paper of Breiman (1995). 6 Breiman’s paper was the inspiration for Tibshirani’s 1996 lasso paper. © 2015 by Taylor & Francis Group, LLC THE NONNEGATIVE GARROTE 21 Lin 2007c, Zou 2006) have shown that the nonnegative garrote has attractive properties when we use other initial estimators such as the lasso, ridge regression or the elastic net. 5 Comparison of the shrinkage behavior of the lasso and the nonnegative garrote for a single variable. Since their λs are on different scales, we used 2 for the lasso and 7 for the garrote to make them somewhat comparable.

On the other hand, the convergence of the parameter estimates from the nonnegative garrote tends to be slower than that of the initial estimate. 21) in the case of an orthonormal model matrix X. 9 q q Estimator 0 Best subset 1 Lasso 2 Ridge Formula √ β˜j · I[|β˜j | > 2λ] sign(β˜j )(|β˜j | − λ)+ β˜j /(1 + λ) Penalties and Bayes Estimates For a fixed real number q ≥ 0, consider the criterion   p p  1 N  2 q minimize (y − x β ) + λ |β | . 21) This is the lasso for q = 1 and ridge regression for q = 2.

Best-subset selection applies the√hard thresholding operator: it leaves the coefficient alone if it is bigger than 2λ, and otherwise sets it to zero. The lasso is special in that the choice q = 1 is the smallest value of q (closest to best-subset) that leads to a convex constraint region and hence a © 2015 by Taylor & Francis Group, LLC SOME PERSPECTIVE 23 convex optimization problem. In this sense, it is the closest convex relaxation of the best-subset selection problem. There is also a Bayesian view of these estimators.

Download PDF sample

Rated 4.75 of 5 – based on 28 votes