-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathreproduce_fig1_fig4_fig5.m
211 lines (152 loc) · 6 KB
/
reproduce_fig1_fig4_fig5.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
clear all
close all
clc
% Paths to access functions from other folders
function_paths = [genpath('tpls/'), genpath('util/'), ...
genpath('predictive_error/'), genpath('baselines/')];
% Add the paths
addpath(function_paths)
% Main Script ==========================================================
% Settings
var_y = 1; % Observation noise Variance
ps = 5; % Number of 0s in theta
K = 12; % Number of available features
var_features = 1; % Range of input data H
var_theta = 1; % Variance of theta
T = 200; % Number of data points
p = K - ps; % True model dimension
% SYNTHETIC DATA =================================================================
%Create data
[y, H, theta] = generate_data(T, K, var_features, var_theta, ps, var_y);
% Indices of true features
idx_h = find(theta ~= 0)';
% Pad original true indices for comparison later
idx_h_padded = [idx_h zeros(1, K - length(idx_h))];
% OLASSO params
epsilon = 1e-7;
% Initial batch of datad
t0 = K+1;
% rjMCMC params
n = round(0.2*T);
Ns = 700;
Nb = 50;
%% CALL METHODS
% TPLS =================================================================
[theta_tpls, idx_tpls, J, plot_stats, idx_tpls_store] = tpls(y, H, K, var_y, t0, idx_h);
% Results for plotting
[tpls_correct, tpls_incorrect] = plot_stats{:};
J_tpls = J;
% OlinLASSO =================================================================
[theta_olin, idx_olin, J, plot_stats, idx_olin_store] = olasso(y, H, t0, epsilon, var_y, idx_h);
% Results for plotting
[olin_correct, olin_incorrect] = plot_stats{:};
J_olin = J;
% OCCD =================================================================
[theta_occd, idx_occd, J, plot_stats, idx_occd_store] = occd(y, H, t0, var_y, idx_h);
% Results for plotting
[occd_correct, occd_incorrect] = plot_stats{:};
J_occd = J;
% RJ MCMC =================================================================
% Data partition and Number of sweeps
[idx_mcmc, theta_RJ, plot_stats, J] = rj_mcmc(y, H, n, Ns, Nb, idx_h, var_y, t0);
% Results for plotting
[mcmc_correct, mcmc_incorrect] = plot_stats{:};
J_mcmc = J;
%% GROUND TRUTHS ===========================================================
% GENIE
[J_true, ~] = true_PE(y, H, t0, T, idx_h, var_y);
% SUPER GENIE
e_super = y(t0+1:end) - H(t0+1:end,:)*theta;
J_super = cumsum(e_super.^2);
% SINGLE EXPECTATIONS =====================================================
[E_add, E_rmv] = expectations(y, H, t0, T, idx_h, var_y, theta);
% BARS (for statistical performance)
tpls_features = [tpls_correct; tpls_incorrect];
olin_features = [olin_correct; olin_incorrect];
occd_features = [occd_correct; occd_incorrect];
mcmc_features = [mcmc_correct; mcmc_incorrect];
%% FIGURE 1: TPLS vs OLinLASSO vs RJMCMC barplots
% Specific run with feature bar plots
% Colors, FontSizes, Linewidths
load plot_settings.mat
fszl= 10;
% Time range to plot
time_plot = t0+1:T;
% BAR PLOTS SPECIFIC RUN =========================================
figure('Renderer', 'painters', 'Position', [900 100 1000 900])
% TPLS
subplot(3,2,1)
formats = {fsz, fszl, fsz_title, lwdt, c_tpls, c_inc, c_true, 'TPLS', 'Time'};
bar_plots(tpls_features, t0+1, T, p, K, formats)
% OLinLASSO
subplot(3,2,2)
formats = {fsz, fszl, fsz_title, lwdt, c_olin, c_inc, c_true, 'OLinLASSO', 'Time'};
bar_plots(olin_features, t0+1, T, p, K, formats)
% OCCD
subplot(3,2,3)
formats = {fsz, fszl, fsz_title, lwdt, c_occd, c_inc, c_true, 'OCCD-TWL', 'Time'};
bar_plots(occd_features, t0+1, T, p, K, formats)
% RJMCMC
subplot(3,2,4)
formats = {fsz, fszl, fsz_title, lwdt, c_mcmc, c_inc, c_true, 'RJMCMC', 'Iteration'};
bar_plots(mcmc_features, 1, Ns, p, K, formats)
% PREDICTIVE ERRORs ===============================================
% Difference
subplot(3,2,6)
hold on
plot(time_plot, J_occd - J_true, 'Color', c_occd, 'LineWidth', lwd)
plot(time_plot, J_olin - J_true, 'Color', c_olin, 'LineWidth', lwd)
plot(time_plot, J_mcmc - J_true, 'Color', c_mcmc, 'LineWidth', lwd)
plot(time_plot, J_tpls - J_true, 'Color', c_tpls, 'LineWidth', lwd)
yline(0, 'Color',c_true, 'linewidth', 2)
hold off
xlim([t0+1, T])
set(gca, 'FontSize', 15)
title('Relative', 'FontSize', fsz_title)
legend('\Delta J_{OCCD}','\Delta J_{OLin}', '\Delta J_{RJMCMC}', '\Delta J_{TPLS}', 'FontSize', fszl-2)
xlabel('Time', 'FontSize', fsz)
ylabel('Predictive Error Difference', 'FontSize', fsz)
grid on
box on
% Raw
subplot(3,2,5)
hold on
plot(time_plot, J_occd, 'Color', c_occd, 'LineWidth', lwd)
plot(time_plot, J_olin, 'Color', c_olin, 'LineWidth', lwd)
plot(time_plot, J_mcmc, 'Color', c_mcmc, 'LineWidth', lwd)
plot(time_plot, J_tpls, 'Color', c_tpls, 'LineWidth', lwd)
plot(time_plot, J_true, 'Color', c_true, 'LineWidth', lwd)
plot(time_plot, J_super, 'Color', c_true, 'LineWidth', lwd, 'LineStyle','--')
hold off
xlim([t0+1, T])
set(gca, 'FontSize', 15)
legend('J_{OCCD}','J_{OLin}', 'J_{RJMCMC}', 'J_{TPLS}', 'J_{GENIE}', 'J_{TRUTH}', 'FontSize', fszl-2)
title('Predictive Error', 'FontSize', fsz_title)
ylabel('Predictive Error ', 'FontSize', fsz)
xlabel('Time', 'FontSize', fsz)
grid on
box on
lwd = lwd+0.5;
%% FIGURES 4 and 5: PLOT EXPECTATIONS
% Import colors
load colors.mat
col{7} = [0,0.5,0];
title_str = 'INSTANT';
y_str = '\Delta_n';
time_plot = time_plot(1:100);
% ADD A FEATURE =======================================================
figure('Renderer', 'painters', 'Position', [0 100 450 900])
subplot(4,1,1)
formats = {fsz, lwd, col, 'Extra Feature', '\Delta^{+m}_n', ''};
expectation_plots(E_add(time_plot,:), time_plot, K-p, formats)
subplot(4,1,2)
formats = {fsz, lwd, col, '', '\Sigma \Delta^{+m}_n', ''};
expectation_plots(cumsum(E_add(time_plot,:)), time_plot, K-p, formats)
% REMOVING A FEATURE ==================================================
%col{7} = [0,0,0];
subplot(4,1,3)
formats = {fsz, lwd, col, 'Removed Feature', '\Delta^{-m}_n', ''};
expectation_plots(E_rmv(time_plot,:), time_plot, p, formats)
subplot(4,1,4)
formats = {fsz, lwd, col, '', '\Sigma \Delta^{-m}_n', 'n'};
expectation_plots(cumsum(E_rmv(time_plot,:)), time_plot, p, formats)