def read_csv(filename):
    f = open(filename)
    lines = f.readlines()
    x = list() # Input list
    y = list() # Output list
    for line in lines:
        temp = line.strip().split(",")
        x.append([float(temp[0].replace('"', ''))])
        y.append([float(temp[1].replace('"', ''))])
    return np.array(x), np.array(y)

def regression():
    iterations = 1000
    degree = 3
    lr = 0.1
    train_test = 0.8
    weights = np.zeros((degree + 1, 1))
    inp, out = read_csv(filename)

    s = np.arange(inp.shape[0])
    np.random.shuffle(s)

    inp = inp[s]
    out = out[s]
    train_inp = inp[:int(train_test*inp.shape[0])]
    train_out = out[:int(train_test*out.shape[0])]
    hx = np.array([[train_inp[i][0] ** j for j in range(degree + 1)] for i in range(train_inp.shape[0])])
    for iteration in range(iterations):
        y = np.dot(hx, weights)
        err = train_out - y
        temp = np.dot(np.transpose(hx), err)
        weights = weights - 2 * lr * temp
        print weights
        print "__-__-__"