Mentions légales du service

Skip to content
Snippets Groups Projects
Commit 28a9b481 authored by Adrien Leman's avatar Adrien Leman Committed by hhakim
Browse files

doc doxygen

parent ebf58863
No related branches found
No related tags found
No related merge requests found
......@@ -10,117 +10,13 @@ FAµST is developped at <a href="http://www.inria.fr/en/centre/rennes"> Rennes I
For more information on the FAµST Project, please visit the <a href="http://faust.gforge.inria.fr"> website</a>. <br>
<HR>
\section install_sec Installation
\subsection step1 Step 1: Download FAµST
FAµST Project is available on subversion repository.
- svn co ./FAUST <br>
\subsection step2 Step 2: Dependency
- library OpenBLAS http://www.openblas.net/
- library HDF5 https://www.hdfgroup.org/HDF5/release/obtainsrc.html#conf
- library MATIO is an C library for reading and writing MATLAB MAT files. https://sourceforge.net/projects/matio/ <br>
matio must be build with hdf5 : Support for MAT file version 7.3 requires the HDF5 library <br>
./configure --enable-extended-sparse=yes --with-matlab=matlab --with-hdf5=/home/library/hdf5-1.8.16/src/.libs <br>
make <br>
make check <br>
- MATLAB Version 2014b <br>
- GCC 4.7 Comptatible with GCC matlab for mex tools <br>
\subsection step3 Step 3: Configure
- ./cmake ../.. <br>
[option] : <br>
-G "CodeBlocks - Unix Makefiles" --> indicates that create both makefile and codeBlock projets. <br>
-DCMAKE_BUILD_TYPE="Debug" --> indicates that compilation is in debug mode. <br>
\subsection step4 Step 4: Make
- make -j4 --> build with 4 threads <br>
- build from codeBlock --> In debug mode for example <br>
\subsection step5 Step 5: Make install
- make install --> copy binary, libraries and documentation in the default output path.
<HR>
\section Example_sec Target & Example
Here is some examples of algorithm of the project FAUST. <br>
\subsection Target1 Target 1 : hierarchicalFactorization_FPP
./build/demo/run_test/src/hierarchicalFactorization_FPP.cpp <br>
Run the hierarchical factorization from a dense matrix contained in a file (can be MEG, Hierfact_compare or HADAMARD or other matrix; cf. directory ./misc/data/mat/ for data example). <br>
The precision FPP (Floating Point Precision) can be double or float. <br>
target name : build/demo/run_test/bin/hierarchicalFactorization_FPP <br>
\subsection Target2 Target 2 : hierarchicalFactorization_gpu_FPP
build/demo/run_test/src/hierarchicalFactorization_gpu_FPP.cpp <br>
Run the hierarchical factorization from a dense matrix contained in a file (can be MEG or other matrix). <br>
The precision FPP (Floating Point Precision) can be double or float. <br>
target name : build/demo/run_test/bin/hierarchicalFactorization_gpu_FPP <br>
# - Compute the multiplication between an faust-matrix and a vector. <br>
# - Evaluation of the time difference using faust-matrix and classical matrix (apply on MEG experiment or dictionnary learning). <br>
#target name : testing/bin/faust_test <br>
\subsection Target3 Target 3 : test_palm4MSA_FPP
./build/demo/run_test/src/test_palm4MSA_FPP.cpp <br>
Run a test of palm4MSA <br>
The precision FPP (Floating Point Precision) can be double or float. <br>
target name : build/demo/run_test/bin/test_palm4MSA_FPP <br>
or ./build/FAUST_OUTPUT/bin/test_palm4MSA_FPP<br>
\subsection Target4 Target 4 : multiply_compare_time_FPP
palm4MSA_test.cpp <br>
Use this script to make runtime comparison between Faust-vector product and Dense matrix-vector product , this script takes two arguments : <br>
- inputfile Faust_example.mat where all the Fausts with different Density, Dimensions and number of factor are stored.<br>
This input file is generated by test/gen_artificial_faust.m (default input file /home/aleman/WORK/FAUST/faust_v1.10/misc/data/mat/Faust_example.mat)<br>
- parameter NB_run the number multiplication made per given configuration (dimension, number of factor, RCG) (default value 100)<br>
<br>
The different computed times are stored in a single mat file :
- \output matfile /home/aleman/WORK/FAUST/faust_v1.10/build/demo/run_test/output/multiply_compare_time.mat
all the different computed times are stored in this single mat file into several variables
tdense : 4D array where all the dense time multiplication are stored
tfaust : 4D array where all the faust time multiplication are stored
DIMS : vector listing all the different dimension that are stored
RCGS : vector listing all the different (Ratio Complexity Gain (theoritical speed-up for faust multiplication))
NB_FACTS : vector listing all the different number of factor for a each configuration of faust
NB_RUN : number, the number of multiplication per configuration
\subsection Target5 Target 5 : launch_hierarchical_fact
configuration file : cmd_line/cmdline_function/launch_hierarchical_fact.cpp <br>
wrapper/cmd_line/src/launch_hierarchical_fact.cpp <br>
Compute the hierarchical factorization of a given data matrix A in cmdline mode. <br>
target name : wrapper/cmd_line/bin/launch_hierarchical_fact <br>
\subsection Target6 Target 6 : launch_palm4MSA
configuration file : cmd_line/cmdline_function/launch_palm4MSA.cpp <br>
wrapper/cmd_line/src/launch_palm4MSA.cpp <br>
Launch_palm4MSA command-line factorizes an input matrix corresponding to an operator of interest into J sparse factors and converges to a stationary point. <br>
target name : wrapper/cmd_line/bin/launch_palm4MSA <br>
<HR>
\section Demo Demonstration
\subsection Demo1 Demo 1 : Source localization in the context of functional brain imaging :
Faust was used in a experience of source localization in brain image.<br>
Different FAUST approximation of a Magnetoencephalography (MEG) gain matrix "Mi" was computed using hierarchical_fact.<br>
A SparseCoding algorithm was used to solve this source localization problem using the MEG gain matrix or its FAUST approximation.<br>
The following pictures illustrate the different trade-offs between speed-up and error of localization using MEG matrix or FAUST matrix.<br>
\image html MEG_computed_time.jpg "computing time" width=2cm
\image html MEG_distance.jpg "distance between estimated sources and true ones" width=2cm
in progress...<br>
\subsection Demo2 Demo 2 : Image denoising :
The goal of this demo is to enhancement a noisy image.<br>
in progress...<br>
\section install_sec Installation
See document Faust.pdf to install the toolbox FAµST.
<HR>
\authors Adrien Leman, Nicolas Bellot, Thomas Gautrais
\date 03/2016
<HR>
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment