-
Notifications
You must be signed in to change notification settings - Fork 0
/
find_best.m
184 lines (147 loc) · 4.75 KB
/
find_best.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
function [] = find_best()
if ~isdeployed
%for matlab2
addpath(genpath('/home/hayashis/git/encode-dp'));
addpath(genpath('/home/hayashis/git/vistasoft'));
addpath(genpath('/home/hayashis/git/jsonlab'));
%for hpc
addpath(genpath('/N/u/brlife/git/encode-dp'));
addpath(genpath('/N/u/brlife/git/vistasoft'));
addpath(genpath('/N/u/brlife/git/jsonlab'));
end
% load my own config.json
config = loadjson('config.json')
if isfield(config,'dtiinit')
disp('using dtiinit aligned dwi')
dt6 = loadjson(fullfile(config.dtiinit, 'dt6.json'))
dwi = fullfile(config.dtiinit, dt6.files.alignedDwRaw)
end
if isfield(config,'dwi')
disp('using dwi')
dwi = config.dwi
end
%need to use different profile directory to make sure multiple jobs won't share the same directory and crash
profile_dir=fullfile('./profile', int2str(feature('getpid')));
mkdir(profile_dir);
c = parcluster();
c.JobStorageLocation = profile_dir;
pool = parpool(c, config.workers);
%load optimal values of parameters
alpha_f = 0;
[alpha_v, lambda_1, lambda_2, ~] = Get_surface_grid_search('results');
% Fit Full Model using optimal parameters and 100% of gradient directions
tic
disp(['FitFullModel.. (alpha_v, lambda_, lambda_2)=(',num2str(alpha_v),',',num2str(lambda_1),',',num2str(lambda_2),')'])
[fe, results] = FitFullModel(...
dwi, ...
config.track, ...
'bogus', ...
config.L, ...
1, ...
alpha_v, ...
alpha_f, ...
lambda_1, ...
lambda_2);
TimeFullFit = toc/60; % Time in minutes
disp(['Time Fitting Full Model=',num2str(TimeFullFit),'mins'])
%fe will be >2G need to save in matlab format
save('fe_optimal.mat', 'fe', '-v7.3');
info.parameters.opt.alpha_v = alpha_v;
info.parameters.opt.lambda_1 = lambda_1;
info.parameters.opt.lambda_2 = lambda_2;
info.fit = results;
save('info.mat', 'info');
%rmpath(info.repo.encode_local);
%rmpath(info.repo.vistasoft);
delete(pool);
end
%% Function that read all
function [alpha_v_min, lambda_1_min, lambda_2_min, error] = Get_surface_grid_search(dataInputPath)
listing = dir(strcat(dataInputPath,'/alpha*.mat'));
Nfiles = size(listing,1);
for n=1:Nfiles
disp(['reading file ',num2str(n),'/',num2str(Nfiles)])
load(strcat(dataInputPath,'/',listing(n).name));
error.alpha_v(n) = results.alpha_v;
error.lambda_1(n) = results.lambda_1;
error.lambda_2(n) = results.lambda_2;
error.value(n) = results.Error_val;
end
[~, ind] = min(error.value); % search for the minimum validation error
alpha_v_min = error.alpha_v(ind);
lambda_1_min = error.lambda_1(ind);
lambda_2_min = error.lambda_2(ind);
plotOptimizationSurface(error, alpha_v_min, lambda_1_min, lambda_2_min)
end
%%
function [] = plotOptimizationSurface(error, alpha_v, lambda_1, lambda_2)
%% plot alpha_v profile
figure
%set(gcf,'units','points','position',[200,400,800,200])
%hold on
ind2 = find(error.lambda_1 == lambda_1);
ind3 = find(error.lambda_2 == lambda_2);
xs = error.alpha_v(intersect(ind2,ind3));
ys = error.value(intersect(ind2,ind3));
%subplot(1,3,1)
scatter(xs, ys)
xlabel('\alpha_v');
ylabel('Error');
saveas(gcf, 'alpha_v.profile.png');
close all;
plot1 = make_plotly_data(xs, ys, 'alpha_v profile', 'alpha_v', 'Error');
%% plot lambda_1 profile
figure
%set(gcf,'units','points','position',[200,400,800,200])
%hold on
ind1 = find(error.alpha_v == alpha_v);
ind3 = find(error.lambda_2 == lambda_2);
xs = error.lambda_1(intersect(ind1,ind3));
ys = error.value(intersect(ind1,ind3));
%subplot(1,3,2)
scatter(xs, ys)
xlabel('\lambda_1');
ylabel('Error');
saveas(gcf, 'lambda_1.profile.png');
close all;
plot2 = make_plotly_data(xs, ys, 'lambda_1 profile', 'lambda_1', 'Error');
%% plot lambda_1 profile
figure
%set(gcf,'units','points','position',[200,400,800,200])
ind1 = find(error.alpha_v == alpha_v);
ind2 = find(error.lambda_1 == lambda_1);
xs = error.lambda_2(intersect(ind1,ind2));
ys = error.value(intersect(ind1,ind2));
%subplot(1,3,3)
scatter(xs, ys)
xlabel('\lambda_2');
ylabel('Error');
saveas(gcf, 'lambda_2.profile.png');
close all;
plot3 = make_plotly_data(xs, ys, 'lambda_2 profile', 'lambda_2', 'Error');
product_json = {plot1, plot2, plot3};
savejson('brainlife', product_json, 'product.json');
end
%% make plotly plot data
function out = make_plotly_data(xs, ys, plotTitle, xaxisTitle, yaxisTitle)
out = struct;
out.data = struct;
out.layout = struct;
out.type = 'plotly';
out.name = plotTitle;
out.data.x = xs;
out.data.y = ys;
out.data.type = 'scatter';
out.data.mode = 'markers';
out.data.marker = struct;
out.data.marker.symbol = 'circle-open';
out.data.marker.size = 8;
out.data = {out.data};
%out.layout.title = plotTitle;
out.layout.xaxis = struct;
out.layout.xaxis.title = xaxisTitle;
out.layout.xaxis.type = 'linear';
out.layout.yaxis = struct;
out.layout.yaxis.title = yaxisTitle;
out.layout.yaxis.type = 'linear';
end