[Octave] Gradient Descent Multi feature

hypothesis theta의 값을 X * theta로 쉽게 구하기 위해 X = [ones(length(data(:, 1)), 1) data] 로 정의.

(쉽게 말하면 X의 맨 왼쪽에 값이 1인 원소의 컬럼을 추가.)

function cost = costFunctionJ(X, y, theta)

m = length(X(:, 1));

h = X * theta;

h = h .^ 2;

cost = 1 / (2 * m) * sum(h);

function [final_theta, history_theta, history_cost] = gradientDescentMulti(X, y, theta, alpha, iteration)

m = length(X(:, 1));

final_theta = theta;

history_theta = [theta];

history_cost = [costFunctionJ(X, y, theta)];

for iter = 1:iteration

h = X * final_theta;

new_theta = final_theta - alpha * 1 / m * ( transpose(X) * (h - y) );

final_theta = new_theta;

history_theta = [ history_theta final_theta ];

history_cost = [ history_cost costFunctionJ(X, y, final_theta) ];

end