1 | %SAMDISTM Distance matrix based on Spectral Angular Mapper (SAM) |
---|
2 | % distance, which is also the spherical geodesic distance |
---|
3 | % |
---|
4 | % D = SAMDISTM (A,B,R) |
---|
5 | % OR |
---|
6 | % D = SAMDISTM (A,B) |
---|
7 | % OR |
---|
8 | % D = SAMDISTM (A,R) |
---|
9 | % OR |
---|
10 | % D = SAMDISTM (A) |
---|
11 | % |
---|
12 | % INPUT |
---|
13 | % A NxK matrix (dataset) |
---|
14 | % B MxK matrix (dataset) |
---|
15 | % R Radius (optional, default: 1) |
---|
16 | % |
---|
17 | % OUTPUT |
---|
18 | % D NxM dissimilarity matrix (dataset) |
---|
19 | % |
---|
20 | % DESCRIPTION |
---|
21 | % Computes the distance matrix D between two sets of vectors, A and B. |
---|
22 | % Distances between vectors X and Y are computed based on the spherical |
---|
23 | % geodesic formula: |
---|
24 | % D(X,Y) = R arcos (X'Y/R^2) |
---|
25 | % X and Y are normalized to a unit length. |
---|
26 | % |
---|
27 | % If A and B are datasets, then D is a dataset as well with the labels defined |
---|
28 | % by the labels of A and the feature labels defined by the labels of B. If A is |
---|
29 | % not a dataset, but a matrix of doubles, then D is also a matrix of doubles. |
---|
30 | % |
---|
31 | % DEFAULT |
---|
32 | % R = 1 |
---|
33 | % |
---|
34 | % REMARKS |
---|
35 | % A square SAM-distance D(A,A) for a finite set A can be proved to be |
---|
36 | % the l_1-distance (LPDISTM). D(A,A).^{1/2} has a Euclidean behavior, so |
---|
37 | % it can be embedded by PSEM in a Euclidean space. |
---|
38 | % |
---|
39 | |
---|
40 | % SEE ALSO |
---|
41 | % JACSIMDISTM, CORRDISTM, LPDISTM, DISTM |
---|
42 | |
---|
43 | % Copyright: Elzbieta Pekalska, ela.pekalska@googlemail.com |
---|
44 | % Faculty EWI, Delft University of Technology and |
---|
45 | % School of Computer Science, University of Manchester |
---|
46 | |
---|
47 | |
---|
48 | function D = samdistm (A,B,r) |
---|
49 | |
---|
50 | bisa = 0; |
---|
51 | if nargin < 2, |
---|
52 | r = 1; |
---|
53 | B = A; |
---|
54 | bisa = 1; |
---|
55 | else |
---|
56 | if nargin < 3, |
---|
57 | if max (size(B)) == 1, |
---|
58 | r = B; |
---|
59 | B = A; |
---|
60 | bisa = 1; |
---|
61 | else |
---|
62 | r = 1; |
---|
63 | end |
---|
64 | end |
---|
65 | end |
---|
66 | |
---|
67 | if r <= 0, |
---|
68 | error ('The parameter R must be positive.'); |
---|
69 | end |
---|
70 | |
---|
71 | isda = isdataset(A); |
---|
72 | isdb = isdataset(B); |
---|
73 | a = +A; |
---|
74 | b = +B; |
---|
75 | [ra,ca] = size(a); |
---|
76 | [rb,cb] = size(b); |
---|
77 | |
---|
78 | if ca ~= cb, |
---|
79 | error ('The matrices should have the same number of columns.'); |
---|
80 | end |
---|
81 | |
---|
82 | aa = sum(a.*a,2); |
---|
83 | bb = sum(b.*b,2)'; |
---|
84 | D = (a*b') ./sqrt(aa(:,ones(rb,1)) .* bb(ones(ra,1),:)); |
---|
85 | D = r * acos(D/r^2); |
---|
86 | |
---|
87 | % Check numerical inaccuracy |
---|
88 | D (find (D < eps)) = 0; % Make sure that distances are nonnegative |
---|
89 | if bisa, |
---|
90 | D = 0.5*(D+D'); % Make sure that distances are symmetric for D(A,A) |
---|
91 | end |
---|
92 | |
---|
93 | |
---|
94 | % Set object labels and feature labels |
---|
95 | if xor(isda, isdb), |
---|
96 | prwarning(1,'One matrix is a dataset and the other not. ') |
---|
97 | end |
---|
98 | if isda, |
---|
99 | if isdb, |
---|
100 | D = setdata(A,D,getlab(B)); |
---|
101 | else |
---|
102 | D = setdata(A,D); |
---|
103 | end |
---|
104 | D.name = 'Distance matrix'; |
---|
105 | if ~isempty(A.name) |
---|
106 | D.name = [D.name ' for ' A.name]; |
---|
107 | end |
---|
108 | end |
---|
109 | return |
---|