Learning Machine Learning from Andrew Ng course on coursera. Storing stuffs in here.

nextTheta = theta - (alpha/m * sumMass)
% m = size(X)(1)
Linear problem: % sumMass = X' * (X * theta - y)

Logistic problem: % sumMass = X' * (g(X * theta) - y)

In ex4-week5, even though I implemented the a2 incorrectly, it still gave me result somewhat correct. (50% slightly wrong - 50% totally right). Weird huh. (using sigmoidGradient(a2') instead of a2'.(1-a2'))