ml

gistfile1.matlab

X = inputs
y = output
m = length(y);
theta = initial_theta = 0;

sigmoid(z) = 1.0 ./ (1.0 + e.^-z);
h = sigmoid(X * theta);

% gradient descent, here we go find theta ...
alpha = 0.001;
num_iters = 1000000;

for iter = 1:num_iters
  gradient = X' * (h - y) / m;
  theta = theta - alpha * gradient;
end

% cost function to check the hypothesis ... should always decreasing
cost = (-y' * log(h) - (1 - y)' * log(1 - h)) / m;