高斯混合模型里面的参数可以用到混核主元分析里面吗？

小小鲁班 · 发表于 2020-12-24 11:27

EDA365欢迎您登录！

您需要登录才可以下载或查看，没有帐号？注册

x

高斯混合模型里面的参数可以用到混核主元分析里面吗？有混合概率主元分析的EM算法的程序吗
function ppca_mixture
filename = 'virus3.dat';
T = importdata(filename);
[N, d] = size(T);
M = 3;  % number of ppca analysers considered 考虑的ppca分析仪数量
q = 2;  % dimension of ppcas  ppca的维数
% init
% initializing the posteriors (p(i|t_n), indexed by (n, i). R in text)
%初始化后验(p(i|t_n)，在文本中由(n, i). R索引
classes = [3, 3, 3, 3, 2, 2, 3, 1, 3, 3, 1, 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 1, 1, 1, 3, 3, 3, 3, 2, 2, 2, 2, 2, 2, 2];
for n=1:N
      for i=1:M
         if(i==classes(n))
            posteriors(n, i) = 1;
         else
            posteriors(n, i) = 0;
         end
      end
end
% precision for convergence checking 收敛检验精度
epsilon = 0.01;
entered = false;

% loop
while(true)
      % updating priors (p(i), pi in text) and mean vectors (mu in text)
      % 更新先验(p(i)，文本中的pi)和平均向量(文本中的mu)
      new_priors = 1/N * sum(posteriors);
      new_mus = zeros(M, d);
      for i=1:M
         for n=1:N
            new_mus(i,

= new_mus(i,

+ posteriors(n, i) * T(n,

;
         end
         new_mus(i, :) = new_mus(i, :) / sum(posteriors(:, i));
      end

      % computing covariance matrices (S in text)计算协方差矩阵(文本中的S)
      covariances = cell(M);
      for i=1:M
         covariances{i} = zeros(d, d);
         for n=1:N
            covariances{i} = covariances{i} + posteriors(n, i)*(T(n, :)' - new_mus(i, :)')*(T(n, :)' - new_mus(i, :)')';
         end
         covariances{i} = covariances{i} / (new_priors(i) * N);
      end

      % applying ppca using covariance matrices 使用协方差矩阵应用ppca
      % (Ws are weight matrices; sigmas are variances) (Ws为权重矩阵;(差异)
      new_Ws = cell(M);
      for i=1:M
         [new_Ws{i}, new_sigmas{i}] = ppca_from_covariance(covariances{i}, q);
      end

      % convergence check 收敛性检查
      if(entered && max(abs(new_priors - priors)) < epsilon && max(max(abs(new_mus - mus))) < epsilon && max(max(max(abs(cell2mat(new_Ws) - cell2mat(Ws))))) < epsilon && max(abs(new_sigmas - sigmas)) < epsilon)
         break;
      end

      % replacing old parameter values 替换旧的参数值
      priors = new_priors;
      mus = new_mus;
      Ws = new_Ws;
      sigmas = new_sigmas;

      % computing the likelihoods (p(t_n|i)) 计算概率(p(t_n|i))
      for i=1:M
         C = sigmas{i}^2 * eye(d) + Ws{i}*Ws{i}';
         detC = det(C);
         invC = inv(C);
         for n=1:N
            likelihoods(n, i) = (2*pi)^(-d/2)*detC^(-1/2)*exp(-1/2*(T(n, :)' - mus(i, :)')'*invC*(T(n, :)' - mus(i, :)'));
         end
      end
      % computing the joint probabilities (p(t_n, i)) 计算联合概率(p(t_n, i))
      for i=1:M
         joint(:, i) = likelihoods(:, i) * priors(i);
      end
      % computing the data priors (p(t_n)) 计算数据先验(p(t_n))
      data_priors = sum(joint');
      % computing and displaying the likelihood of the data set 计算和显示数据集的可能性
      disp(sum(log(data_priors)));
      % computing the posteriors (p(i|t_n), indexed by (n, i). R in text)
      % 计算后验(p(i|t_n)，用(n, i)索引。R在文本中)
      for n=1:N
         posteriors(n, :) = joint(n, :) / data_priors(n);
      end

      % we went through the loop at least once 我们至少做了一次循环
      entered = true;
end

% computing the latent variables 计算潜变量
latent = cell(M);
for i=1:M
      latent{i} = ppca_latent(T, Ws{i}, sigmas(i));
end
% selecting likely points for each ppca analyser 为每个ppca分析器选择可能的点
for n=1:N
      [~, i] = sort(posteriors(n, :), 'descend');
      points_to_classes(n) = i(1);
end
classes_to_points = cell(M);
classes_to_point_numbers = cell(M);
for n=1:N
      classes_to_points{points_to_classes(n)}(:, end+1) = latent{points_to_classes(n)}(:, n);
      classes_to_point_numbers{points_to_classes(n)}(end+1) = n;
end

% display the results of the automatic classification 显示自动分类结果
for i=1:M
      disp(classes_to_point_numbers{i});
end

ppca_plot2d(classes_to_points{1});
end

zaiyiaaaa · 发表于 2020-12-24 13:35

kekek · 发表于 2020-12-24 13:28

zzz.dan · 发表于 2020-12-24 13:13

帮你顶一下

帐号		自动登录	找回密码
密码			注册

高斯混合模型里面的参数可以用到混核主元分析里面吗？

EDA365欢迎您登录！

浏览过的版块

推荐内容 /1