Coursera机器学习 week4 assignment

lrCostFunction.m:

b = sigmoid(X*theta);J = -(1/m)*( (log(b'))*y + log(1-b')*(1-y) ) + (lambda/(2*m))*(sum(theta.^2)-theta(1)^2);k = X'*(b-y);grad(1) = (1/m)*k(1);long = length(k);k2 = (1/m) * (k(2:long,:)) + (lambda/m)*(theta(2:long,:));grad(2:long,:) = k2;

oneVsAll.m:

initial_theta = zeros(n + 1, 1);options = optimset('GradObj', 'on', 'MaxIter', 50);for i = 1:num_labels<pre name="code" class="plain">z = X*all_theta';   %5000*10[a, b] = max(z,[],2);p = b;

[theta] = ... fmincg (@(t)(lrCostFunction(t, X, (y==i), lambda)), ... initial_theta, options); all_theta(i,:) = theta';end

predictOneVsAll.m:

z = X*all_theta';   %5000*10[a, b] = max(z,[],2);p = b;

predict.m:

X = [ones(m, 1) X];   %(5000*401)X2 = sigmoid(X*Theta1');   %(5000*401) (25*401)  ——>5000*25X2 = [ones(m, 1) X2];X3 = sigmoid(X2*Theta2');   %(5000*10)[a, b] = max(X3,[],2);p = b;