[coursera]Machine Learning(Andrew Ng)部分编程题答案(三)

Andrew Ng在coursera上的机器学习公开课 自己写的部分编程答案，留作记录，欢迎交流

sigmoid.m

a=ones(size(z));
g=a./(a.+exp(-z));

costFunction.m

s=sigmoid(sum(theta'.*X,2));
 J=-1/m* sum((  y.*log(s)+ (ones(size(y)) -y).*log(ones(size(s))-s)),1);
grad=1/m*  sum((s-y).*X,1)';

costFunctionReg.m

theta2=theta;
theta2(1,1)=0;
s=sigmoid(sum(theta'.*X,2));
J=-1/m* sum((  y.*log(s)+ (ones(size(y)) -y).*log(ones(size(s))-s)),1)+ lambda/(2*m)*sum(theta2.^2);
grad=1/m*  sum((s-y).*X,1)' + lambda/m*theta2;

predict.m

s=sigmoid(sum(theta'.*X,2));
for i =1:m
    if s(i,1) >=0.5
        p(i, 1)=1;
    else
        p(i , 1)=0;
    endif
endfor