Mentions légales du service

Skip to content
Snippets Groups Projects
Commit 2b9a9536 authored by Adrien Leman's avatar Adrien Leman Committed by hhakim
Browse files

readme

parent 561af552
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 91 additions and 49 deletions
README.txt
+ 91
− 49
View file @ 2b9a9536
......@@ -3,7 +3,6 @@
######## Flexible Approximate Multi-Layer Sparse Transform ############
##########################################################################
##########################################################################
General purpose:
......@@ -12,13 +11,10 @@ designed to factorize matrices of interest into multiple sparse factors.
It contains a template CPU/GPU C++ code and a Matlab wrapper.
The algorithms implemented here are described in details in [1]- Le Magoarou
For more information on the FAuST Project, please visit the website of the
project: <http://faust.gforge.inria.fr>
##########################################################################
##########################################################################
License:
......@@ -41,13 +37,9 @@ You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
##########################################################################
##########################################################################
Installation:
Please refer to the document Faust.PDF, for more detailed informations
about the installation.
Quick install :
1- Unpack the directory.
2- mkdir ./build
......@@ -59,69 +51,116 @@ Quick install :
5- make
6- make install
Please refer to the document Faust.PDF, for more detailed informations
about the installation.
Warning :
mex function working with specific version of gcc depend to the platform used.
For that, please refers to the matworks information:
http://fr.mathworks.com/support/compilers/R2016a/index.html
Windows install :
Warning :
for Matlab installation, function "where" doesn't work when Path name have
space caracter like "Program file". please prefers a witout space path like "programfiles"
where gcc // where g++ // where cmake // where make must be available.
For windows platform :
cmake -G "MinGW Makefiles" ..
message(STATUS "CXX=$ENV{CXX}")
sous windows on utilise MinGW pour le compilateur GCC G++ au lieu de visual studio plous lourd à installer. MinGW version 4.9.2 pour les mex functions..
set(CTEST_CMAKE_GENERATOR "MinGW Makefiles")
To compile MEX functions with Matlab:
You can use the MinGW-w64 version 4.9.2 compiler from TDM-GCC to build MEX files and standalone MATLAB® engine and MAT-file applications.
creer un compte mathworks et installer depuis la plateforme MinGW-w64 version 4.9.2 compiler from TDM-GCC
(cf. http://fr.mathworks.com/help/matlab/matlab_external/install-mingw-support-package.html for more information)
##########################################################################
##########################################################################
Main tools :
The main functions offered by the FAuST toolbox are implemented in two files
- hierarchical_fact.m
[lambda, facts, errors] = hierarchical_fact(params);
input:
- params: Structure contained data matrice, desired number
of factors, constraint sets...
output:
- lambda: multiplicative scalar
- facts : estimated factorization (cell array of sparse matrices)
- errors
- palm4MSA.m
[lambda, facts] = palm4MSA(params)
input:
- params: Structure contained data matrice, desired number
of factors, constraint sets...
output:
- lambda: multiplicative scalar
- facts : estimated factorization (cell array of sparse matrices)
The main tools offered by the FAuST toolbox are implemented in two mex function:
[lambda, facts, errors] = mexHierarchical_fact(params)
[lambda, facts] = mexPalm4MSA(params)
These functions are available in the wrapper matlab.
###### [lambda, facts, errors] = mexHierarchical_fact(params) #######################
# This mex function runs the hierarchical
# matrix factorization algorithm (Algorithm 3 of [1])on the specified
# input matrix, returning the factors in "facts" (cell array of sparse matrices),
# the multiplicative scalar in "lambda" and the errors in "errors".
#
# Required fields in PARAMS:
# --------------------------
# 'data' - Training data.
# A matrix to hierarchically factorize.
# 'nfacts' - Number of factors.
# Specifies the desired number of factors.
# 'cons' - Constraint sets.
# Specifies the constraint sets in which eafaust_params_palm.ch factor should lie. It
# should be a cell-array of size 2*(nfacts-1), where the jth columns
# sets the constraints for the jth factorization in two factors:
# cons(1,j) specifies the constraints for the left factor and
# cons(2,j) for the right factor. cons(i,j) should be itself a
# cell-array of size 1*4 taking this form for a factor of size m*n:
# {'constraint name', 'constraint parameter', m, n}
#
# Optional fields in PARAMS:
# --------------------------
# 'niter1' - Number of iterations for the 2-factorisation.
# Specifies the desired number of iteration for the factorisations in
# 2 factors. The default value is 500.
# 'niter2' - Number of iterations for the global optimisation.
# Specifies the desired number of iteration for the global
# optimisation. The default value is 500.
# 'fact_side' - Side to be factorized iteratively: 1-the left or
# 0-the right. The default value is 0;
# 'verbose' - Verbosity of the function. if verbose=1, the function
# outputs the error at each iteration. if verbose=0, the function runs
# in silent mode. The default value is 0.
# 'update_way' - Way in which the factors are updated. If update_way = 1
# ,the factors are updated from right to left, and if update_way = 0,
# the factors are updated from left to right. The default value is 0.
###### [lambda, facts] = mexPalm4MSA(params)
# Factorization of a data matrix into multiple factors using PALM.
# This mex function runs the PALM algorithm on the
# specified set of signals (Algorithm 2 of [1]), returning the factors in
# "facts" and the multiplicative scalar in "lambda".
#
# Required fields in PARAMS:
# --------------------------
# 'data' - Training data.
# A matrix containing the training signals as its columns.
# 'nfacts' - Number of factors.
# Specifies the desired number of factors.
# 'cons' - Constraint sets.
# Specifies the constraint sets in which each factor should lie. It
# should be a cell-array of size 1*nfacts, where the jth column sets
# the constraints for the jth factor (starting from the left). cons(j)
# should be itself a cell-array of size 1*4 taking this form for a
# factor of size m*n:
# {'constraint name', 'constraint parameter', m, n}
# 'niter' - Number of iterations.
# Specifies the number of iterations to run.
# 'init_facts' - Initialization of "facts".
# Specifies a starting point for the algorithm.
# Optional fields in PARAMS:
# --------------------------
# 'init_lambda' - Initialization of "lambda".
# Specifies a starting point for the algorithm. The default value is 1
# 'verbose' - Verbosity of the function. if verbose=1, the function
# outputs the error at each iteration. if verbose=0, the function runs
# in silent mode. The default value is 0.
# 'update_way' - Way in which the factors are updated. If update_way = 1
# ,the factors are updated from right to left, and if update_way = 0,
# the factors are updated from left to right. The default value is 0.
##########################################################################
class FAUST :
It represents a given dense matrix by a product of sparse matrix (i.e faust)
in order to speed-up multiplication by this matrix.
##########################################################################
##########################################################################
Demonstration:
This package contains the Matlab code which represented the results of [1].
One demonstration is available in "./misc/demo/" directory:
- Source_localization using MEG data (cf. Sec.V. of [1];)
See related ./misc/demo/README file for more detail.
One demonstration is available in "./DIR_INSTALL/demo/" directory:
- Brain_source_localization using MEG data (cf. Sec.V. of [1];)
- Hadamard_factorization
- Runtime_comparison
This package contains a CTest tool to execute binary file automatically.
(cf. CMakeLists.txt)
##########################################################################
##########################################################################
Naming conventions in the FAuST project :
Naming conventions in the C++ FAuST project :
- namespace : Faust::xxx
- class : Faust::MyClass / Faust::Class
......@@ -147,6 +186,9 @@ Naming conventions in the FAuST project :
Contacts:
Luc Le Magoarou: luc.le-magoarou@inria.fr
Remi Gribonval : remi.gribonval@inria.fr
Nicolas Bellot : nicolas.bellot@inria.fr
Adrien Leman : adrien.leman@inria.fr
Thomas Gautrais : thomas.gautrais@inria.fr
##########################################################################
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment