ex3练习

• 逻辑回归—多分类

  % 载入数据，自动保存在数组X, y
  load('ex3data1.mat');
  m = size(X, 1);
  
  % 更新theta，多分类all_theta为num_labels*(n+1)，二分类theta为(n+1)*1
  lambda = 0.1;
  [all_theta] = oneVsAll(X, y, num_labels, lambda);
  
  % 训练预测准确率
  pred = predictOneVsAll(all_theta, X);
  fprintf('\nTraining Set Accuracy: %f\n', mean(double(pred == y)) * 100);

  % 代价函数和偏导，同二分类，有正则项
  theta_cp = theta;
  theta_cp(1) = 0;
  J = (1/m)*((-y)'*log(sigmoid(X*theta))-(1-y)'*log(1-sigmoid(X*theta)))+(lambda/(2*m))*(theta_cp'*theta_cp);
  grad = (1/m)*(X'*(sigmoid(X*theta)-y))+(lambda/m)*theta_cp;

  % 多分类
  n = size(X, 2);   % n个特征
  all_theta = zeros(num_labels, n + 1);
  initial_theta = zeros(n + 1, 1);
  options = optimset('GradObj', 'on', 'MaxIter', 50);
  
  % 训练num_labels，得到每类的最佳theta,存放于一行
  for i=1:num_labels
  	[all_theta(i,:)] = fmincg (@(t)(lrCostFunction(t, X, (y == i), lambda)),initial_theta, options);
  end

  % 预测,max找到每行最大值
  [p_value,p] = max(sigmoid(X*(all_theta')),[],2);

   

   

   

• 神经网络

  % 载入数据
  load('ex3data1.mat');
  m = size(X, 1);
  
  % 载入theta
  load('ex3weights.mat');
  
  %训练预测准确率
  pred = predict(Theta1, Theta2, X);
  fprintf('\nTraining Set Accuracy: %f\n', mean(double(pred == y)) * 100);

  % 预测
  X = [ones(m,1) X];      % 特征矩阵
  X_hid = sigmoid(X*Theta1');       % 隐藏层激活值，Theta1:25*401,25隐藏层单元数，400输入层单元数
  m_hid = size(X_hid,1);
  X_hid = [ones(m_hid,1) X_hid];     % 增加a_0 = 1
  h_theta = sigmoid(X_hid*Theta2');  % 输出值，Theta2:10*26,10输出层单元数，25隐藏层单元数
  [p_value,p] = max(h_theta,[],2);