Function gradient

05_1_gradientdescent_manually.py:24–25

(x, y, y_pred)

# J = MSE = 1/N * (w*x - y)**2
# dJ/dw = 1/N * 2x(w*x - y)
def gradient(x, y, y_pred):
    return np.mean(2*x*(y_pred - y))

print(f'Prediction before training: f(5) = {forward(5):.3f}')