h = sigmoid(X * theta);
f = -y.*log(h);
s = (1-y).*log(1-h);
J = (f-s);
grad = X'*(h - y);
b = lambda/(2*m);
c = theta .^ 2;
J = a*sum(J)+b*sum(c);
temp = theta;
temp(1) = 0;
grad = a*sum(grad) + sum(temp*lambda);