Back
Projected Newton-type methods in machine learning
{We consider projected Newton-type methods for solving large-scale optimization problems arising in machine learning and related fields. We first introduce an algorithmic framework for projected Newton-type methods by reviewing a canonical projected (quasi-)Newton method. This method, while conceptually pleasing, has a high computation cost per iteration. Thus, we discuss two variants that are more scalable, namely, two-metric projection and inexact projection methods. Finally, we show how to apply the Newton-type framework to handle non-smooth objectives. Examples are provided throughout the chapter to illustrate machine learning applications of our framework.}
@incollection{escidoc:0279, title = {{Projected Newton-type methods in machine learning}}, booktitle = {{Optimization for Machine Learning}}, abstract = {{We consider projected Newton-type methods for solving large-scale optimization problems arising in machine learning and related fields. We first introduce an algorithmic framework for projected Newton-type methods by reviewing a canonical projected (quasi-)Newton method. This method, while conceptually pleasing, has a high computation cost per iteration. Thus, we discuss two variants that are more scalable, namely, two-metric projection and inexact projection methods. Finally, we show how to apply the Newton-type framework to handle non-smooth objectives. Examples are provided throughout the chapter to illustrate machine learning applications of our framework.}}, pages = {305--330}, publisher = {MIT Press}, address = {Cambridge, MA, USA}, year = {2011}, slug = {escidoc-0279}, author = {Schmidt, M. and Kim, D. and Sra, S.}, url = {http://www.kyb.tuebingen.mpg.de//fileadmin/user\textunderscoreupload/files/publications/2011\textunderscoreOPT\textunderscoreChapter\textunderscore6824[0].pdf} }