-
Notifications
You must be signed in to change notification settings - Fork 1
/
Fig_07b.m
240 lines (203 loc) · 9.49 KB
/
Fig_07b.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
% Figure002.m
% Algorithm performance vs compression ratio
clearvars;
filename = 'Figure002';
ts_disp([filename ' started running.']);
%% Parameters
% Search parameters
% redo_search = [0,0,0,0]; load([filename '.mat']); % comment if full search required
% Default parameters
set_default_parameters;
% Trial Parameters
set_trial_parameters;
% Solvers to run
solvers = {'RPCA','RPCA+L1-DF','RPCA+BOT-DF','RPCA+UOT-DF'};
%% Run trials
if on_cluster, PaceParalleltoolbox_r2016b(true); end
for i = 1:length(M_space)
% Update trial variable
sim_param.M = M_space(i);
% Parameter search (efficiently using parallelization)
if ~exist('redo_search')
ts_disp('Running full parameter search.');
solver_param = get_solver_param_via_search(solvers,search_seeds,F1_threshold,eval_weights,sim_param,ADMM_opts);
else
for s = 1:length(redo_search)
if redo_search(s)
ts_disp(['Running parameter search for ' solvers{s} '.']);
solver_param(s) = get_solver_param_via_search(cellstr(solvers{s}),search_seeds,F1_threshold,eval_weights,sim_param,ADMM_opts);
else
ts_disp(['Retrieving parameters from ' solvers{s} ' : '...
'lambda = ' num2str(results(s,i,1).param.lambda) ', '...
'gamma = ' num2str(results(s,i,1).param.gamma) ', '...
'kappa = ' num2str(results(s,i,1).param.kappa) ', '...
'mu = ' num2str(results(s,i,1).param.mu) ...
]);
solver_param(s) = results(s,i,1).param;
end
end
end
% Discretize the solvers and trials space (for parallelization)
[t_grid,s_grid] = meshgrid( 1:nbr_trials , 1:length(solvers) );
% Run trials (efficiently using parallelization)
ts_disp(['Running parallel trials...']);
S_parfor = zeros(sim_param.imsize(1)*sim_param.imsize(2), sim_param.nbr_frames, numel(s_grid));
L_parfor = zeros(sim_param.imsize(1)*sim_param.imsize(2), sim_param.nbr_frames, numel(s_grid));
time_parfor = zeros(numel(s_grid),1);
parfor j = 1:numel(s_grid)
s = s_grid(j); t = t_grid(j);
% Simulate Problem
rng(t,'twister'); [Y,Phi] = simulator(sim_param);
% Run solver
[S_parfor(:,:,j),L_parfor(:,:,j),time_parfor(j)] = ...
run_solver(solvers{s},sim_param.imsize,Y,Phi,solver_param(s),ADMM_opts);
end
% Save metrics
time = zeros(length(solvers),nbr_trials);
F1 = zeros(length(solvers),nbr_trials);
S_rMSE = zeros(length(solvers),nbr_trials);
L_rMSE = zeros(length(solvers),nbr_trials);
for j = 1:numel(s_grid)
s = s_grid(j); t = t_grid(j);
% Simulate Problem
rng(t,'twister'); [~,~,S_gt,L_gt] = simulator(sim_param);
% Save data
results(s,i,t).S = S_parfor(:,:,j);
results(s,i,t).L = L_parfor(:,:,j);
results(s,i,t).time = time_parfor(j);
results(s,i,t).solver = solvers{s};
results(s,i,t).param = solver_param(s);
% Evaluation metrics
time(s,t) = time_parfor(j);
F1(s,t) = compute_F1(results(s,i,t).S,S_gt,F1_threshold);
S_rMSE(s,t) = norm(vec(results(s,i,t).S-S_gt))^2 / norm(vec(S_gt))^2;
L_rMSE(s,t) = norm(vec(results(s,i,t).L-L_gt))^2 / norm(vec(L_gt))^2;
% % Display in terminal
% disp([var_name ' = ' num2str(sim_param.noise_sigma) ' ' ...
% 'Trial #' num2str(t) '/' num2str(nbr_trials) '- ' ...
% solvers{s} ' (time=' num2str(results(s,i,t).time) 's) : '...
% 'F1 = ' num2str( F1(s,t) ) ', '...
% 'S_rMSE = ' num2str( S_rMSE(s,t) ) ', '...
% 'L_rMSE = ' num2str( L_rMSE(s,t) ) ]);
end
% Display summmary statistics
ts_disp(['Summary statistics of ' num2str(size(results,3)) ' trials '...
'for ' var_name ' = ' num2str(sim_param.M) ]);
fprintf('Method\t\t\tTime\t\tF1 score\tS_rMSE\t\tL_rMSE\n');
for s = 1:length(solvers)
switch solvers{s}
case 'RPCA', fprintf([solvers{s} '\t\t\t']);
otherwise, fprintf([solvers{s} '\t\t']);
end
fprintf('%08.4f\t',median(time(s,:)));
fprintf('%08.4f\t',mean(F1(s,:)));
fprintf('%08.4f\t',mean(S_rMSE(s,:)));
fprintf('%08.4f\t\n',mean(L_rMSE(s,:)));
end
% Display Progress
ts_disp(['Progress of ' filename ' : ' num2str(i/length(M_space)*100) '% completed.']);
end
% Save Data
clear S_parfor L_parfor time_parfor S_gt L_gt F1 S_rMSE L_rMSE time
close all; save(filename);
if no_display, return; end
%% Generate figure
% Compute metrics
S_rMSE = nan(length(solvers),length(M_space),nbr_trials);
L_rMSE = nan(length(solvers),length(M_space),nbr_trials);
F1 = nan(length(solvers),length(M_space),nbr_trials);
runtime = nan(length(solvers),length(M_space),nbr_trials);
for i = 1:length(M_space)
sim_param.M = M_space(i); % Update trial variable
for t = 1:nbr_trials
for s = 1:length(solvers)
% Simulate Problem
rng(t,'twister'); [~,~,S_gt,L_gt] = simulator(sim_param);
% Evaluation metrics
compute_S_rMSE = @(S) norm(vec(S-S_gt))^2 / norm(vec(S_gt))^2;
compute_L_rMSE = @(L) norm(vec(L-L_gt))^2 / norm(vec(L_gt))^2;
compute_F1_score = @(S) compute_F1(S,S_gt,F1_threshold);
% Compute
S_rMSE(s,i,t) = compute_S_rMSE(results(s,i,t).S);
L_rMSE(s,i,t) = compute_L_rMSE(results(s,i,t).L);
F1(s,i,t) = compute_F1_score(results(s,i,t).S);
runtime(s,i,t) = results(s,i,t).time;
end
end
end
% Plot options
fontSize = 14;
markers = {'o','s','d','^'};
cmap = [ 0 0.4470 0.7410;
0.8500 0.3250 0.0980;
0.9290 0.6940 0.1250;
0.4940 0.1840 0.5560;
0.4660 0.6740 0.1880;
0.3010 0.7450 0.9330;
0.6350 0.0780 0.1840];
% Plot
fig = figure(1); clf;
set(fig,'Units','normalized','Position',[0.1 0.0 0.3 1.0]); clf;
S_rMSE_lines = []; L_rMSE_lines = []; F1_lines = []; runtime_lines = [];
for s = 1:length(solvers)
subplot(411);
% h(1,s).mainLine = errorbar(M_space,mean(squeeze(rMSE(s,:,:))'),std(squeeze(rMSE(s,:,:))'),...
% 'LineWidth',2,'Color',cmap(s,:),'Marker',markers{s});
h(1,s) = shadedErrorBar(M_space,squeeze(S_rMSE(s,:,:))',{@nanmean,@nanstd},...
'lineprops',{['-' markers{s}],'LineWidth',2,'Color',cmap(s,:),'MarkerFaceColor',cmap(s,:)}); hold on;
S_rMSE_lines = [S_rMSE_lines, h(1,s).mainLine];
hold on;
subplot(412);
% h(1,s).mainLine = errorbar(M_space,mean(squeeze(rMSE(s,:,:))'),std(squeeze(rMSE(s,:,:))'),...
% 'LineWidth',2,'Color',cmap(s,:),'Marker',markers{s});
h(2,s) = shadedErrorBar(M_space,squeeze(L_rMSE(s,:,:))',{@nanmean,@nanstd},...
'lineprops',{['-' markers{s}],'LineWidth',2,'Color',cmap(s,:),'MarkerFaceColor',cmap(s,:)}); hold on;
L_rMSE_lines = [L_rMSE_lines, h(2,s).mainLine];
hold on;
subplot(413);
% h(2,s).mainLine = errorbar(M_space,mean(squeeze(F1(s,:,:))'),std(squeeze(F1(s,:,:))'),...
% 'LineWidth',2,'Color',cmap(s,:),'Marker',markers{s});
h(3,s) = shadedErrorBar(M_space,squeeze(F1(s,:,:))',{@nanmean,@nanstd},...
'lineprops',{['-' markers{s}],'LineWidth',2,'Color',cmap(s,:),'MarkerFaceColor',cmap(s,:)}); hold on;
F1_lines = [F1_lines, h(3,s).mainLine];
hold on;
subplot(414);
% h(2,s).mainLine = errorbar(M_space,mean(squeeze(runtime(s,:,:))'),std(squeeze(F1(s,:,:))'),...
% 'LineWidth',2,'Color',cmap(s,:),'Marker',markers{s});
h(3,s) = shadedErrorBar(M_space,squeeze(runtime(s,:,:))',{@nanmean,@nanstd},...
'lineprops',{['-' markers{s}],'LineWidth',2,'Color',cmap(s,:),'MarkerFaceColor',cmap(s,:)}); hold on;
runtime_lines = [runtime_lines, h(3,s).mainLine];
hold on;
end
subplot(411);
% loglog(M_space,mean(rMSE,3)','LineWidth',2);
axis tight; grid on; grid minor; ylim([0,1]);
% set(gca, 'XScale', 'log', 'YScale', 'log'); ylim([0,1]);
% set(gca, 'XScale', 'log');
set(gca, 'YScale', 'log');
% xlabel(var_name,'FontSize',fontSize);
ylabel('Signal error','FontSize',fontSize);
% legend(S_rMSE_lines,solvers,'FontSize',fontSize,'Location','SouthWest');
subplot(412);
% loglog(M_space,mean(rMSE,3)','LineWidth',2);
axis tight; grid on; grid minor; ylim([0,0.2]);
% set(gca, 'XScale', 'log', 'YScale', 'log'); ylim([0,1]);
% set(gca, 'XScale', 'log');
% xlabel(var_name,'FontSize',fontSize);
ylabel('Low rank error','FontSize',fontSize);
% legend(L_rMSE_lines,solvers,'FontSize',fontSize,'Location','NorthWest');
subplot(413);
%loglog(M_space,mean(F1,3)','LineWidth',2);
axis tight; grid on; grid minor; ylim([0,1]);
% set(gca, 'XScale', 'log');
% xlabel(var_name,'FontSize',fontSize);
ylabel('Support estimation','FontSize',fontSize);
% legend(F1_lines,solvers,'FontSize',fontSize,'Location','SouthWest');
subplot(414);
%loglog(M_space,mean(runtime,3)','LineWidth',2);
axis tight; grid on; grid minor; %ylim([0,1]);
% set(gca, 'XScale', 'log');
xlabel(var_name,'FontSize',fontSize);
ylabel('Run time (s)','FontSize',fontSize);
% legend(F1_lines,solvers,'FontSize',fontSize,'Location','SouthWest');
drawnow; saveFig2PDF(filename);