-
Notifications
You must be signed in to change notification settings - Fork 0
/
WS_transpositions.m
135 lines (120 loc) · 4.63 KB
/
WS_transpositions.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
function [transStat, elapsedTimes] = WS_transpositions(lossMtx,nGroup1, nGroup2, nTrans, permNo)
% function [transStat, elapsedTimes] = transposition_test(lossMtx,nGroup1, nGroup2, nTrans, permNo)
%
% This function performs the transposition procedure by transposing group labels and
% summing pair-wise distances from a pre-computed loss matrix. The transposition test
% for ratio statistic is given in
%
% [1] Songdechakraiwut, T., Shen, L., Chung, M.K. 2021 Topological learning and
%its application to multimodal brain network integration, Medical Image
%Computing and Computer Assisted Intervention (MICCAI), LNCS 12902:166-176
%
% [2] Songdechakraiwut, T. Chung, M.K. 2023 Topological learning for brain networks,
% Annals of Applied Statistics 17:403-433, arXiv:2012.00675.
% https://arxiv.org/pdf/2012.00675.pdf
%
%
% INPUT
% lossMtx : loss matrix whose entries are all possible pair-wise distances
% nGroup1 : sample size of group 1
% nGroup2 : sample size of group 2
% nTrans : the number of transpositions being performed
% permNo : intermix random permutation every 'permNo' transpositions
%
%
% OUTPUT
% transStat : array of ratio statistic, 1 x nTrans
% elapsedTimes : elapsed time, 1 x nTrans
%
% %If you are using any part of the code, please reference one of the above paper.
%
% (C) 2020 Tananun Songdechakraiwut, Moo K. Chung
% University of Wisconsin-Madison
%
% Contact tananun@cs.wisc.edu or mkchung@wisc.edu
% for support/permission with the codes
%
% Update history
% 2020 November 11, created - Tananun Songdechakraiwut
% 2020 December 10, edited Moo Chung
%
tStart = tic;
totalNo = nGroup1 + nGroup2;
% denomGroup1 = (nGroup1*(nGroup1-1) + nGroup2*(nGroup2-1))/2;
% denomGroup2 = nGroup1*nGroup2;
% constant = denomGroup1 / denomGroup2;
transStat = zeros(1, nTrans);
elapsedTimes = zeros(1, nTrans);
for t = 1:round(nTrans/permNo)
startInd = (t - 1) * permNo + 1;
%% Initialize sum of pair-wise distances for both within and between groups
permutation = randperm(totalNo);
% dividing group labels into two groups
permutedG1 = permutation(1:nGroup1);
permutedG2 = permutation(nGroup1+1:totalNo);
% within groups
within = 0;
% sum of pair-wise distances within groups
for i = 1:nGroup1 % group 1
for j = i + 1:nGroup1
within = within + lossMtx(permutedG1(i), permutedG1(j));
end
end
for i = 1:nGroup2 % group 2
for j = i + 1:nGroup2
within = within + lossMtx(permutedG2(i), permutedG2(j));
end
end
% between groups
% sum of pair-wise distances between groups
between = 0;
for i = 1:nGroup1
for j = 1:nGroup2
between = between + lossMtx(permutedG1(i), permutedG2(j));
end
end
% saving the ratio statistics
transStat(startInd) = between/within; % * constant;
elapsedTimes(startInd) = toc(tStart);
%% transposition procedure
prevWithin = within;
prevBetween = between;
for n = startInd + 1:startInd + permNo - 1
% random transposition indices
ind1 = randi(nGroup1);
ind2 = randi(nGroup2);
% within groups
within = 0;
% sum of offset distances being removed from within-group distances
for i = 1:nGroup1 % group 1
within = within + lossMtx(permutedG1(i), permutedG1(ind1));
end
for i = 1:nGroup2 % group 2
within = within + lossMtx(permutedG2(i), permutedG2(ind2));
end
% remove excess distances to itself
within = within - lossMtx(permutedG1(ind1), permutedG1(ind1));
within = within - lossMtx(permutedG2(ind2), permutedG2(ind2));
% between groups
% sum of offset distances being removed from between-group distances
between = 0;
for i = 1:nGroup2
between = between + lossMtx(permutedG1(ind1), permutedG2(i));
end
for i = 1:nGroup1
between = between + lossMtx(permutedG1(i), permutedG2(ind2));
end
between = between - 2 * lossMtx(permutedG1(ind1), permutedG2(ind2));
% update iteratively
delta = between - within;
prevWithin = prevWithin + delta;
prevBetween = prevBetween - delta;
% saving the ratio statistics
transStat(n) = prevBetween/prevWithin; % * constant;
elapsedTimes(n) = toc(tStart);
% swapping networks corresponding to indices
temp = permutedG1(ind1);
permutedG1(ind1) = permutedG2(ind2);
permutedG2(ind2) = temp;
end
end