# Talk:Matrix Completion.m

### From Wikimization

(Difference between revisions)

Current revision (15:55, 11 December 2009) (edit) (undo)m (Protected "Talk:Matrix Completion.m" [edit=autoconfirmed:move=autoconfirmed]) |
|||

(3 intermediate revisions not shown.) | |||

Line 1: | Line 1: | ||

- | Most of the subroutines are part of the [http://svt.caltech.edu SVT package] coded by Emmanuel Candes and Stephen Becker. The package is still in beta, and is being updated frequently. The package also contains an implementation of FPC applied to the matrix completion problem | + | Most of the subroutines are part of the [http://svt.caltech.edu SVT package] coded by Emmanuel Candes and Stephen Becker. The package is still in beta, and is being updated frequently. The package also contains an implementation of FPC applied to the matrix completion problem (see [http://convexoptimization.com/TOOLS/Goldfarb.pdf Goldfarb's paper]); the implementation departs from Goldfarb's implementation by using an iterative SVD solver (the same solver that SVT uses), instead of a randomized SVD solver. SVT and FPC often complement each other; on problems when one works poorly, then other method might work well. Please contact me if you have any questions. -Stephen [[User:Srbecker|Srbecker]] 19:22, 23 March 2009 (PDT) |

- | (see [http://convexoptimization.com/TOOLS/Goldfarb.pdf Goldfarb's paper]); the implementation departs from Goldfarb's implementation by using an iterative SVD solver (the same solver that SVT uses), instead of a randomized SVD solver. SVT and FPC often complement each other; on problems when one works poorly, then other method might work well. Please contact me if you have any questions. -Stephen [[User:Srbecker|Srbecker]] 19:22, 23 March 2009 (PDT) | + |

## Current revision

Most of the subroutines are part of the SVT package coded by Emmanuel Candes and Stephen Becker. The package is still in beta, and is being updated frequently. The package also contains an implementation of FPC applied to the matrix completion problem (see Goldfarb's paper); the implementation departs from Goldfarb's implementation by using an iterative SVD solver (the same solver that SVT uses), instead of a randomized SVD solver. SVT and FPC often complement each other; on problems when one works poorly, then other method might work well. Please contact me if you have any questions. -Stephen Srbecker 19:22, 23 March 2009 (PDT)