-
Notifications
You must be signed in to change notification settings - Fork 4
/
nk_MultiPerfComp.m
55 lines (49 loc) · 1.81 KB
/
nk_MultiPerfComp.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
function GDanalysis = nk_MultiPerfComp(GDanalysis, multi_pred, label, ngroups, act)
if ~exist('act','var') || isempty(act), act = 'pred'; end
switch act
case 'pred'
fld = 'MultiClass';
% Compute multi-class probabilties
multi_prob = nk_ConvProbabilities(multi_pred, ngroups);
case 'prob'
fld = 'MultiClassProb';
multi_prob = multi_pred;
end
lx = size(multi_prob,1);
pred = nan(lx,1);
stdpred = pred; ci1 = pred; ci2 = pred;
% Loop through cases
if iscell(multi_pred)
for i=1:lx
if isempty(multi_pred{i})
pred(i) = NaN; stdpred(i) = NaN; ci1(i) = NaN; ci2(i) = NaN; errs(i) = NaN;
else
% Maximum probability decides about multi-class membership
[maximum,pred(i)] = max(multi_prob(i,:),[],'includenan');
if isnan(maximum), pred(i)=NaN; continue; end
% Is this useful: ?
stdpred(i) = std(multi_pred{i});
ci = percentile(multi_pred{i},[2.5 97.5]);
ci1(i) = ci(1); ci2(i) = ci(2);
end
end
else
%numpred = size(multi_pred,2);
[maximum,pred] = max(multi_prob,[],2,'includenan');
inan = isnan(maximum);
pred(inan)=NaN;
stdpred = std(multi_pred,[],2);
stdpred(inan)=NaN;
ci = cell2mat(arrayfun( @(i) percentile(multi_pred(i,:),[2.5 97.5]),1:lx,'UniformOutput',false )');
ci(inan)=NaN;
ci1 = ci(:,1); ci2 = ci(:,2);
end
% Compute multi-class performance
GDanalysis = nk_MultiEvalPerf(GDanalysis, label, pred, ngroups, fld);
% Store prediction results in structure
GDanalysis.(fld).multi_probabilitiesCV2 = multi_prob;
GDanalysis.(fld).multi_predictionsCV2 = pred;
GDanalysis.(fld).multi_predictionsCV2_std = stdpred;
GDanalysis.(fld).multi_predictionsCV2_ci1 = ci1;
GDanalysis.(fld).multi_predictionsCV2_ci2 = ci2;
end