MMML/compute_metric_learning.m at master · GitWR/MMML · GitHub

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
function [ U ] = compute_metric_learning( k_Gras, k_Spd, lamda1, lamda2, Train_lables)
  num_class = length(unique(Train_lables)); %
  D = size(k_Gras,1); %
  d = 8; % the target dimensionality of the generated subspace
  itera = 1;
  % U = rand(D,d);
for i=1:itera
  fprintf('\n i= %d \n',i);

  Sw=zeros(D,D); %
  Sb=zeros(D,D); %

  % compute the within-class scatter matrix
  Nw=0;
  Nb=0;
  for j = 1 : num_class
      num_eachclass = find(Train_lables==j);
      for k = 1 : length(num_eachclass)
          K_gras_data1 = k_Gras(num_eachclass(k),:);
          K_spd_data1 = k_Spd(num_eachclass(k),:);
          for m=k+1 : length(num_eachclass)
              K_gras_data2 = k_Gras(num_eachclass(m),:);
              K_spd_data2 = k_Spd(num_eachclass(m),:);
              Sw_temp_gras = lamda1*(K_gras_data1'-K_gras_data2')*(K_gras_data1'-K_gras_data2')'*lamda1;
              Sw_temp_spd = lamda2*(K_spd_data1'-K_spd_data2')*(K_spd_data1'-K_spd_data2')'*lamda2;
              Sw_temp = Sw_temp_gras + Sw_temp_spd;
              Sw = Sw+Sw_temp;
              Nw=Nw+1;
          end
      end
  end

  % compute the between-class scatter matrix
  for j=1:num_class
      num_eachclass=find(Train_lables==j);
      num_difclass=find(Train_lables~=j);
      for k=1:length(num_eachclass)
          K_gras_data1 = k_Gras(num_eachclass(k),:);
          K_spd_data1 = k_Spd(num_eachclass(k),:);
          for m=1:length(num_difclass)
              K_gras_data2 = k_Gras(num_difclass(m),:);
              K_spd_data2 = k_Spd(num_difclass(m),:);
              Sb_temp_gras = lamda1*(K_gras_data1'-K_gras_data2')*(K_gras_data1'-K_gras_data2')'*lamda1;
              Sb_temp_spd = lamda2*(K_spd_data1'-K_spd_data2')*(K_spd_data1'-K_spd_data2')'*lamda2;
              Sb_temp = Sb_temp_gras + Sb_temp_spd;
              Sb=Sb+Sb_temp;
              Nb=Nb+1;%
          end
      end
  end

  Sw_final=Sw/(Nw);
  Sb_final=Sb/(Nb);

  %Object_function = Sw_final-alpha*Sb_final;
  % if (i>1)
  %     if (Cost(i)>=Cost(i-1))
  %         count1=count1+1;
  %         if (count1>=1)
  %            break;
  %         end
  %     end
  % end

  % optimization
  [Object_V , Object_E] = eig(inv(Sw_final)*Sb_final);
  E_unsort = diag(Object_E);
  [~ , index] = sort(E_unsort,'descend');
  V_sort = Object_V(:,index);
  U = V_sort(:,1:d);

  Cost(i) = det(U'*Sb_final*U)/det(U'*Sw_final*U);
  fprintf(' iter\t            cost val\t \n    ');
  fprintf('%5d\t \n%+.16e\t \n', i, Cost(:,1:i));

end
end