Gradient Descent (Linear regression)

gradient_descent.py

import random

X_Y = [(2, 4), (3, 6), (4, 9), (5, 10), (6, 12), (7, 15), (8,17), (9,20)]

w = [1, 2, 3, 4]
a = 0.0001
iterations = 15000
h = lambda x: w[0] * x**3 + w[1] * x**2 + w[2]*x + w[3]
l = lambda x, y: (h(x) - y)**2
dl = lambda x, y, w_: 2.0 * (h(x) - y) * w_


def gradient_descent():
    for _ in range(iterations):
        # TODO: Use dot product to improve performance.
        w[0] = w[0] - a / 2 * sum(dl(*x_i, w[0]) for x_i in X_Y)
        w[1] = w[1] - a / 2 * sum(dl(*x_i, w[1]) for x_i in X_Y)
        w[2] = w[2] - a / 2 * sum(dl(*x_i, w[2]) for x_i in X_Y)
        w[3] = w[3] - a / 2 * sum(dl(*x_i, w[3]) for x_i in X_Y)


gradient_descent()
print("weight", [round(w_, 2) for w_ in w])