This is the GFLseg package for multiple change-point detection in
multi-dimensional signals.
Copyright 2009-2011, Kevin Bleakley and Jean-Philippe Vert.
INTRODUCTION
GFLseg implements the group fused Lasso method for multiple
change-point detections in a multidimensional signal. Two algorithms are
implemented. The first is an exact solver for the group fused Lasso. The
second is the group fused LARS, which is faster but less accurate. In
addition, segmentation by dynamic programming is also implemented. This
is more accurate but much slower. GFLseg contains also functions to
automatically process array CGH data.
INSTALLATION
To run this software you need MATLAB (tested on MATLAB 7,
version R2009b). To use the software, you just need to let MATLAB know
where the functions are. For this, suppose the current file is in the
directory "/GFLseg". You need to start MATLAB and, at the
command prompt type:
>> addpath(genpath('/GFLseg'))
USAGE
If you have a p-dimensional signal of length n in a n*p matrix Y,
the simplest way to use GFLseg is to type at the MATLAB prompt:
>> segmentation = simpleGFL(Y)
It will find candidate change-points in Y with the group Lars method, and
then fine-tune the change-point selection with dynamic programming. The
location of the change-points will be available in the segmentation.jumps
vector. For more details and options, type:
>> help simpleGFL
The simpleGFL function is a wrapper for several functions that perform
specific tasks. You can call also them directly:
gflars.m : segmentation with the group fused LARS for a given number of
breakpoints.
gflasso.m : segmentation with the group fused Lasso for a given set of
regularization parameter(s).
gflassoK.m : segmentation with the group fused Lasso for a given
number of breakpoints.
gflassoKatleast.m : segmentation with the group fused Lasso for at
least a given number of breakpoints.
dpseg.m : segmentation by dynamic programming and automatic selection
of the number of change-points
If you have CGH array data, you may directly type the following to
perform joint segmentation on each chromosome and output a file in seeGH
format:
>> gainlossCGH(clonefile,datafile,outfile)
The "demo/" directory contain a number of demonstration scripts to
illustrate the use of GFLseg. For each function, detailed explanations
and available options are obtained by typing at the Matlab prompt:
>> help
LICENSE
GFLseg is licensed under the GPL version 3 or any later version
(see LICENSE)
REFERENCE
If you use this package please cite the following reference:
K. Bleakley and J.-P. Vert, The group fused Lasso for multiple change-point detection. 2011.
CONTACT
Jean-Philippe.Vert@mines.org
kevbleakley@gmail.com