Mentions légales du service

Skip to content
Snippets Groups Projects
Commit 73d2fb95 authored by hhakim's avatar hhakim
Browse files

Complete the hadamard fact. demos integration into +matfaust by changing... 

Complete the hadamard fact. demos integration into +matfaust by changing filepath for output figures, creating the needed parent directory in run_all_demo.m and removing the original scripts in misc/demo/Hadamard_factorization.

This integration has started with 61390ef0.
parent ba835d3f
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 6 additions and 593 deletions
misc/demo/Hadamard_factorization/demo_fact_hadamard.m deleted 100644 → 0
+ 0
− 163
View file @ ba835d3f
%% Description demo_fact_hadamard
%
% This demo hierarchically factorizes the Hadamard dictionary and then
% plots the results. This essentially reproduces figure 2 from [1].
%
% For more information on the FAuST Project, please visit the website of
% the project : <http://faust.inria.fr>
%
%% License:
% Copyright (2016): Nicolas Bellot, Adrien Leman, Thomas Gautrais, Luc Le Magoarou, Remi Gribonval
% INRIA Rennes, FRANCE
% http://www.inria.fr/
%
% The FAuST Toolbox is distributed under the terms of the GNU Affero
% General Public License.
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU Affero General Public License as published
% by the Free Software Foundation.
%
% This program is distributed in the hope that it will be useful, but
% WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
% See the GNU Affero General Public License for more details.
%
% 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/>.
%
%% Contacts:
% Nicolas Bellot : nicolas.bellot@inria.fr
% Adrien Leman : adrien.leman@inria.fr
% Thomas Gautrais : thomas.gautrais@inria.fr
% Luc Le Magoarou : luc.le-magoarou@inria.fr
% Remi Gribonval : remi.gribonval@inria.fr
%
%% References:
% [1] Le Magoarou L. and Gribonval R., "Flexible multi-layer sparse
% approximations of matrices and applications", Journal of Selected
% Topics in Signal Processing, 2016.
% <https://hal.archives-ouvertes.fr/hal-01167948v1>
%
%%
n = 32;
M = log2(n);
%% Generating the data
[matrix,~] = hadamard_mat(M);
%% Setting of the parameters
params_hadamard.nrow=n;
params_hadamard.ncol=n;
params_hadamard.nfacts = M;
params_hadamard.cons = cell(2,M-1);
for j=1:M-1
params_hadamard.cons{1,j} = {'splincol',2,n,n};
params_hadamard.cons{2,j} = {'splincol',n/2^j,n,n};
end
params_hadamard.niter1 = 30;
params_hadamard.niter2 = 30;
params_hadamard.update_way = 1;
params_hadamard.verbose = 0;
hadamard_faust = faust_decompose(matrix,params_hadamard);
Xhat = full(hadamard_faust);
relative_error = norm(matrix - Xhat)/norm(matrix);
fprintf(['\n\n relative error between hadamard matrix and its transform : ' num2str(relative_error) '\n']);
nb_mult= 500;
dense_times=zeros(nb_mult,1);
faust_times=zeros(nb_mult,1);
for i=1:nb_mult
x=rand(n,1);
y_faust=rand(n,1);
y_X=rand(n,1);
tic
y_X=matrix*x;
t1=toc;
tic
y_faust=hadamard_faust*x;
t2=toc;
dense_times(i)=t1;
faust_times(i)=t2;
end
dense_t=mean(dense_times);
faust_t=mean(faust_times);
speed_up=dense_t/faust_t;
disp(['multiplication by hadamard matrix : ' num2str(dense_t) 'sec']);
disp(['multiplication by faust : ' num2str(faust_t) 'sec']);
disp(['multiplication speed-up using faust : ' num2str(speed_up) ]);
%% Plotting the results
% figure configuration
runPath=which(mfilename);
pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures'];
format_fig='-dpng';
fighandle=figure;
hold on;
subplot(1,params_hadamard.nfacts+1,1);
imagesc(Xhat); axis square
set(gca,'xtick',[],'ytick',[])
%get the factor of the Faust
facts=cell(1,M);
for i=1:M
facts{i}=get_factor(hadamard_faust,i);
end
for kk = 1:params_hadamard.nfacts
subplot(1,params_hadamard.nfacts+1,1+kk)
imagesc(facts{kk}); axis square
set(gca,'xtick',[],'ytick',[])
end
fighandle.Name =['Hadamard-factorisation_image'];
figure_name = [figure_dir filesep 'Hadamard-factorisation_image'];
print(figure_name, format_fig);
fighandle=figure;
hold on;
subplot(1,params_hadamard.nfacts+1,1);
imagesc(Xhat); axis square
set(gca,'xtick',[],'ytick',[])
for kk = 1:params_hadamard.nfacts
subplot(1,params_hadamard.nfacts+1,1+kk)
spy(facts{kk}); axis square
set(gca,'xtick',[],'ytick',[])
end
fighandle.Name =['Hadamard-factorisation_nnz_coeff'];
figure_name = [figure_dir filesep 'Hadamard-factorisation_nnz_coeff'];
print(figure_name, format_fig);
misc/demo/Hadamard_factorization/norm_hadamard.m deleted 100644 → 0
+ 0
− 198
View file @ ba835d3f
%% Description norm_hadamard
%
% This demo makes some time comparison between the 2-norm of the Hadamard matrix and
% her Faust representation for different dimension
% of the Hadamard matrix.
%
% For more information on the FAuST Project, please visit the website of
% the project : <http://faust.inria.fr>
%
%% License:
% Copyright (2016): Nicolas Bellot, Adrien Leman, Thomas Gautrais, Luc Le Magoarou, Remi Gribonval
% INRIA Rennes, FRANCE
% http://www.inria.fr/
%
% The FAuST Toolbox is distributed under the terms of the GNU Affero
% General Public License.
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU Affero General Public License as published
% by the Free Software Foundation.
%
% This program is distributed in the hope that it will be useful, but
% WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
% See the GNU Affero General Public License for more details.
%
% 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/>.
%
%% Contacts:
% Nicolas Bellot : nicolas.bellot@inria.fr
% Adrien Leman : adrien.leman@inria.fr
% Thomas Gautrais : thomas.gautrais@inria.fr
% Luc Le Magoarou : luc.le-magoarou@inria.fr
% Remi Gribonval : remi.gribonval@inria.fr
%
%% References:
% [1] Le Magoarou L. and Gribonval R., "Flexible multi-layer sparse
% approximations of matrices and applications", Journal of Selected
% Topics in Signal Processing, 2016.
% <https://hal.archives-ouvertes.fr/hal-01167948v1>
%%
import matfaust.Faust
nb_mult=10;
Ms=6:11;
ns=2.^Ms;
nb_dim=length(Ms);
threshold=10^(-10);
h = waitbar(0,'speed up hadamard : Generation of the data ...');
Hadamard_matrices=cell(1,nb_dim);
Hadamard_facts=cell(1,nb_dim);
for k=1:nb_dim
waitbar(k/nb_dim);
M=Ms(k);
n=ns(k);
% generation of the hadamard factorisation
[H,facts] = hadamard_mat(M);
Hadamard_matrices{k}=H;
Hadamard_facts{k}=facts;
end
close(h);
dense_times=zeros(nb_mult,nb_dim);
faust_times=zeros(nb_mult,nb_dim);
norm_dense=zeros(1,nb_dim);
norm_faust=zeros(1,nb_dim);
%RCGs=ns./(Ms*2);
RCGs=zeros(1,nb_dim);
h = waitbar(0,'2-norm hadamard : multiplication time comparison ...');
for i=1:nb_mult
waitbar(i/nb_mult);
for k=1:nb_dim
n=ns(k);
hadamard_dense=Hadamard_matrices{k};
hadamard_faust=Faust(Hadamard_facts{k});
RCGs(k)=rcg(hadamard_faust);
%% 2-norm of the hadamard matrix
tic
norm_dense(k)=norm(hadamard_dense);
t1=toc;
tic
norm_faust(k)=norm(hadamard_faust);
t2=toc;
dense_times(i,k)=t1;
faust_times(i,k)=t2;
end
end
close(h);
mean_dense_t = mean(dense_times);
mean_faust_t = mean(faust_times);
speed_up = mean_dense_t ./ mean_faust_t;
% expected_norm = ones(1,nb_dim); % version orthonormee
expected_norm = 2.^(Ms/2); % version non orthonormée
err_norm_dense=sqrt((norm_dense - expected_norm).^2);
err_norm_faust=sqrt((norm_faust - expected_norm).^2);
disp(['speed-up : ' num2str(speed_up)]);
disp(['norm dense : ' num2str(norm_dense(1,:))]);
disp(['norm faust : ' num2str(norm_faust(1,:))]);
%% Plot the result
plot_tickness=2.0;
legend_location = 'NorthWest';
f=figure;
% runtime comparison
subplot(1,3,1);
semilogy(Ms,mean_faust_t,'linewidth',plot_tickness);
hold on
semilogy(Ms,mean_dense_t,'r','linewidth',plot_tickness);
ymin=min([mean_dense_t,mean_faust_t]);
ymax=max([mean_dense_t,mean_faust_t]);
grid on
axis([Ms(1) Ms(end) ymin ymax]);
legend('faust','dense','Location',legend_location);
ylabel('Computed Time (sec)');
xlabel('log(dim)');
title('runtime ');
set(gca,'XTick',Ms);
% speed-up
subplot(1,3,2);
semilogy(Ms,speed_up,'linewidth',plot_tickness);
hold on
semilogy(Ms,ones(1,nb_dim),'k','linewidth',plot_tickness);
semilogy(Ms,RCGs,'g','linewidth',plot_tickness);
grid on
axis([Ms(1) Ms(end) min([speed_up,1,RCGs]) max([speed_up,1,RCGs])]);
title('speed-up norm(A)');
xlabel('log(dim)');
ylabel('speedup');
legend('faust','neutral','theoretical','Location',legend_location);
set(gca,'XTick',Ms);
%
subplot(1,3,3);
id=find(err_norm_faust ~= 0);% semilogy is not compatible with 0
semilogy(Ms(id),err_norm_faust(id),'r+-','linewidth',plot_tickness);
hold on
semilogy(Ms,err_norm_dense,'linewidth',plot_tickness);
ymin=min([err_norm_dense, err_norm_faust]);
ymax=max([err_norm_dense, err_norm_faust]);
grid on
%axis([Ms(1) Ms(end) ymin ymax]);
legend('faust','dense','Location',legend_location);
ylabel('Computed Time (sec)');
xlabel('log(dim)');
title('error ');
set(gca,'XTick',Ms);
f.Name =['Hadamard 2-norm'];
%% save the figure
runPath=which(mfilename);
pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures'];
format_fig='-dpng';
figure_name=[figure_dir filesep 'Hadamard-norm'];
print(figure_name, format_fig);
misc/demo/Hadamard_factorization/speed_up_hadamard.m deleted 100644 → 0
+ 0
− 228
View file @ ba835d3f
%% Description speed_up_hadamard
%
% This demo makes some time comparison between (Hadamard matrix)-vector multiplication and
% (Hadamard factorisation i.e a FAµST)-vector multiplication for different dimension
% of the Hadamard matrix.
%
% For more information on the FAuST Project, please visit the website of
% the project : <http://faust.inria.fr>
%
%% License:
% Copyright (2016): Nicolas Bellot, Adrien Leman, Thomas Gautrais, Luc Le Magoarou, Remi Gribonval
% INRIA Rennes, FRANCE
% http://www.inria.fr/
%
% The FAuST Toolbox is distributed under the terms of the GNU Affero
% General Public License.
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU Affero General Public License as published
% by the Free Software Foundation.
%
% This program is distributed in the hope that it will be useful, but
% WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
% See the GNU Affero General Public License for more details.
%
% 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/>.
%
%% Contacts:
% Nicolas Bellot : nicolas.bellot@inria.fr
% Adrien Leman : adrien.leman@inria.fr
% Thomas Gautrais : thomas.gautrais@inria.fr
% Luc Le Magoarou : luc.le-magoarou@inria.fr
% Remi Gribonval : remi.gribonval@inria.fr
%
%% References:
% [1] Le Magoarou L. and Gribonval R., "Flexible multi-layer sparse
% approximations of matrices and applications", Journal of Selected
% Topics in Signal Processing, 2016.
% <https://hal.archives-ouvertes.fr/hal-01167948v1>
%%
import matfaust.Faust
nb_mult=500;
Ms=6:12;
ns=2.^Ms;
nb_dim=length(Ms);
threshold=10^(-10);
h = waitbar(0,'speed up hadamard : Generation of the data ...');
Hadamard_matrices=cell(1,nb_dim);
Hadamard_facts=cell(1,nb_dim);
for k=1:nb_dim
waitbar(k/nb_dim);
M=Ms(k);
n=ns(k);
% generation of the hadamard factorisation
[H,facts] = hadamard_mat(M);
Hadamard_matrices{k}=H;
Hadamard_facts{k}=facts;
end
close(h);
dense_times=zeros(nb_mult,nb_dim);
faust_times=zeros(nb_mult,nb_dim);
dense_trans_times=zeros(nb_mult,nb_dim);
faust_trans_times=zeros(nb_mult,nb_dim);
faust_mtimes_trans_times=zeros(nb_mult,nb_dim);
h = waitbar(0,'speed up hadamard : multiplication time comparison ...');
for i=1:nb_mult
waitbar(i/nb_mult);
for k=1:nb_dim
n=ns(k);
hadamard_dense=Hadamard_matrices{k};
hadamard_faust=Faust(Hadamard_facts{k});
x=rand(n,1);
ydense=zeros(n,1);
yfaust=zeros(n,1);
ydense_trans=zeros(n,1);
yfaust_trans=zeros(n,1);
yfaust_mtimes_trans=zeros(n,1);
%% multiplication by the hadamard matrix
tic
ydense=hadamard_dense*x;
t1=toc;
tic
yfaust=hadamard_faust*x;
t2=toc;
if(norm(ydense-yfaust)>threshold)
error('speed_up hadamard : multiplication problem');
end
%% multiplication by the transpose of the hadamard matrix
tic
ydense_trans=hadamard_dense'*x;
t3=toc;
tic
yfaust_trans=hadamard_faust'*x;
t4=toc;
tic
yfaust_mtimes_trans=mtimes_trans(hadamard_faust,x,1);
t5=toc;
tic
if (norm(ydense_trans - ydense_trans) > threshold)
error('speed_up hadamard : transpose multiplication problem');
end
if (yfaust_trans ~= yfaust_mtimes_trans)
error('speed_up hadamard : transpose multiplication problem');
end
dense_times(i,k)=t1;
faust_times(i,k)=t2;
dense_trans_times(i,k)=t3;
faust_trans_times(i,k)=t4;
faust_mtimes_trans_times(i,k)=t5;
end
end
close(h);
mean_dense_t = mean(dense_times);
mean_faust_t = mean(faust_times);
speed_up = mean_dense_t ./ mean_faust_t;
mean_dense_trans_t = mean(dense_trans_times);
mean_faust_trans_t = mean(faust_trans_times);
mean_faust_mtimes_trans_t = mean(faust_mtimes_trans_times);
speed_up_trans = mean_dense_trans_t ./ mean_faust_trans_t;
speed_up_mtimes_trans = mean_dense_trans_t ./ mean_faust_mtimes_trans_t;
%% Plot the result
plot_tickness=2.0;
legend_location = 'NorthWest';
f=figure;
subplot(2,2,1);
semilogy(Ms,mean_faust_t,'linewidth',plot_tickness);
hold on
semilogy(Ms,mean_dense_t,'r','linewidth',plot_tickness);
ymin=min([mean_dense_t,mean_faust_t]);
ymax=max([mean_dense_t,mean_faust_t]);
grid on
axis([Ms(1) Ms(end) ymin ymax]);
legend('faust','dense','Location',legend_location);
ylabel('Computed Time (sec)');
xlabel('log(dim)');
title('runtime Hadamard A*x');
set(gca,'XTick',Ms);
subplot(2,2,2);
semilogy(Ms,speed_up,'linewidth',plot_tickness);
hold on
semilogy(Ms,ones(1,nb_dim),'k','linewidth',plot_tickness);
grid on
axis([Ms(1) Ms(end) min([speed_up,1]) max([speed_up,1])]);
title('speed-up Hadamard A*x');
xlabel('log(dim)');
ylabel('speedup');
legend('faust','neutral','Location',legend_location);
set(gca,'XTick',Ms);
subplot(2,2,3);
semilogy(Ms,mean_faust_trans_t,'linewidth',plot_tickness);
hold on
semilogy(Ms,mean_dense_trans_t,'r','linewidth',plot_tickness);
semilogy(Ms,mean_faust_mtimes_trans_t,'g','linewidth',plot_tickness);
ymin=min([mean_dense_trans_t,mean_faust_trans_t,mean_faust_mtimes_trans_t]);
ymax=max([mean_dense_trans_t,mean_faust_trans_t,mean_faust_mtimes_trans_t]);
grid on
axis([Ms(1) Ms(end) ymin ymax]);
legend('faust','dense','mtimes_trans','Location',legend_location);
ylabel('Computed Time (sec)');
xlabel('log(dim)');
title('runtime Hadamard A''*x');
set(gca,'XTick',Ms);
subplot(2,2,4);
semilogy(Ms,speed_up_trans,'linewidth',plot_tickness);
hold on
semilogy(Ms,ones(1,nb_dim),'k','linewidth',plot_tickness);
semilogy(Ms,speed_up_mtimes_trans,'g','linewidth',plot_tickness);
grid on
axis([Ms(1) Ms(end) min([speed_up_trans,1,speed_up_mtimes_trans]) max([speed_up_trans,1,speed_up_mtimes_trans])]);
title('speed-up Hadamard A''*x');
xlabel('log(dim)');
ylabel('speedup');
legend('faust','neutral','mtimes_trans','Location',legend_location);
set(gca,'XTick',Ms);
f.Name =['Hadamard Faust-vector multiplication'];
%% save the figure
runPath=which(mfilename);
pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures'];
format_fig='-dpng';
figure_name=[figure_dir filesep 'Hadamard-speed_up'];
print(figure_name, format_fig);
misc/demo/run_all_demo.m
+ 3
− 1
View file @ 73d2fb95
...@@ -44,6 +44,7 @@ ...@@ -44,6 +44,7 @@
%% %%
%% Quick start %% Quick start
mkdir Figures
disp('*********** Quick Start Demos *************'); disp('*********** Quick Start Demos *************');
quick_start; quick_start;
factorize_matrix; factorize_matrix;
...@@ -60,7 +61,8 @@ Fig_BSL; ...@@ -60,7 +61,8 @@ Fig_BSL;
%% Hadamard factorization %% Hadamard factorization
disp('*********** Hadamard Factorization *************'); disp('*********** Hadamard Factorization *************');
demo_fact_hadamard; import matfaust.demo.hadamardfact.*
demo_fact_hadamard;
speed_up_hadamard; speed_up_hadamard;
norm_hadamard; norm_hadamard;
......
wrapper/matlab/+matfaust/+demo/+hadamardfact/demo_fact_hadamard.m
+ 1
− 1
View file @ 73d2fb95
...@@ -124,7 +124,7 @@ function demo_fact_hadamard() ...@@ -124,7 +124,7 @@ function demo_fact_hadamard()
% figure configuration % figure configuration
runPath=which(mfilename); runPath=which(mfilename);
pathname = fileparts(runPath); pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures']; figure_dir = [ '.' filesep 'Figures'];
format_fig='-dpng'; format_fig='-dpng';
......
wrapper/matlab/+matfaust/+demo/+hadamardfact/norm_hadamard.m
+ 1
− 1
View file @ 73d2fb95
...@@ -191,7 +191,7 @@ function norm_hadamard() ...@@ -191,7 +191,7 @@ function norm_hadamard()
%% save the figure %% save the figure
runPath=which(mfilename); runPath=which(mfilename);
pathname = fileparts(runPath); pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures']; figure_dir = ['.' filesep 'Figures'];
format_fig='-dpng'; format_fig='-dpng';
figure_name=[figure_dir filesep 'Hadamard-norm']; figure_name=[figure_dir filesep 'Hadamard-norm'];
print(figure_name, format_fig); print(figure_name, format_fig);
......
wrapper/matlab/+matfaust/+demo/+hadamardfact/speed_up_hadamard.m
+ 1
− 1
View file @ 73d2fb95
...@@ -224,7 +224,7 @@ function speed_up_hadamard() ...@@ -224,7 +224,7 @@ function speed_up_hadamard()
%% save the figure %% save the figure
runPath=which(mfilename); runPath=which(mfilename);
pathname = fileparts(runPath); pathname = fileparts(runPath);
figure_dir = [pathname filesep '..' filesep 'Figures']; figure_dir = [ '.' filesep 'Figures'];
format_fig='-dpng'; format_fig='-dpng';
figure_name=[figure_dir filesep 'Hadamard-speed_up']; figure_name=[figure_dir filesep 'Hadamard-speed_up'];
print(figure_name, format_fig); print(figure_name, format_fig);
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment