AdaptiveModel-basedTransfer/ProbabilityModel.m at master · BingshuiDa/AdaptiveModel-basedTransfer · 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
classdef ProbabilityModel % Works reliably for 2(+) Dimensional distributions
  properties
    modeltype; % multivariate normal ('mvarnorm' - for real coded) or univariate marginal distribution ('umd' - for binary coded)
    mean_noisy;
    mean_true;
    covarmat_noisy;
    covarmat_true;
    probofone_noisy;
    probofone_true;
    probofzero_noisy;
    probofzero_true;
    vars;
  end
  methods (Static)
    function model = ProbabilityModel(type)
      model.modeltype = type;
    end
    function solutions = sample(model,nos)
      if strcmp(model.modeltype,'mvarnorm')
        solutions = mvnrnd(model.mean_true,model.covarmat_true,nos);
      elseif strcmp(model.modeltype,'umd')
        solutions = rand(nos,model.vars);
        for i = 1:nos
            index1 = solutions(i,:) <= model.probofone_true;
            index0 = solutions(i,:) > model.probofone_true;
            solutions(i,index1) = 1;
            solutions(i,index0) = 0;
        end
      end
    end
    function probofsols = pdfeval(model,solutions)
      if strcmp(model.modeltype,'mvarnorm')
        probofsols = mvnpdf(solutions,model.mean_noisy,model.covarmat_noisy);
      elseif strcmp(model.modeltype,'umd')
        nos = size(solutions,1);
        probofsols = zeros(nos,1);
        probvector = zeros(1,model.vars);
        for i = 1:nos
          index = solutions(i,:) == 1;
          probvector(index) = model.probofone_noisy(index);
          index = solutions(i,:) == 0;
          probvector(index) = model.probofzero_noisy(index);
          probofsols(i) = prod(probvector);
        end
      end
    end
    function model = buildmodel(model,solutions)
      [pop,model.vars] = size(solutions);
      if strcmp(model.modeltype,'mvarnorm')
        model.mean_true = mean(solutions);
        covariance = cov(solutions);
        model.covarmat_true = diag(diag(covariance)); % Simplifying to univariate distribution by ignoring off diagonal terms of covariance matrix
        solutions_noisy = [solutions;rand(round(0.1*pop),model.vars)];
        model.mean_noisy = mean(solutions_noisy);
        covariance = cov(solutions_noisy);
        model.covarmat_noisy = diag(diag(covariance));% Simplifying to univariate distribution by ignoring off diagonal terms of covariance matrix
        model.covarmat_noisy = cov(solutions_noisy);
      elseif strcmp(model.modeltype,'umd')
        model.probofone_true = mean(solutions);
        model.probofzero_true = 1 - model.probofone_true;
        solutions_noisy = [solutions;round(rand(round(0.1*pop),model.vars))];
        model.probofone_noisy = mean(solutions_noisy);
        model.probofzero_noisy = 1 - model.probofone_noisy;
      end
    end
  end
end