def estimate_beta(x, y):
beta_initial = [random.random() for x_i in x]
return minimize_stochastic(squared_error,
squared_error_gradient,
x, y,
beta_initial,
0.001)
def squared_error_gradient(x_i, y_i, theta):
alpha, beta = theta
return [-2 * error(alpha, beta, x_i, y_i), # alpha partial derivative
-2 * error(alpha, beta, x_i, y_i) * x_i] # beta partial derivative
# choose random value to start
random.seed(0)
theta = [random.random(), random.random()]
alpha, beta = minimize_stochastic(squared_error,
squared_error_gradient,
num_friends_good,
daily_minutes_good,
theta,
0.0001
)
return alpha, beta
import sys
sys.path.append(
'/Users/rileylittlefield/Desktop/notes/readingnotes/python-ml/data-science-from-scratch/09-exercises'
)
from simple_linear_regression import error
import random
from pokemon_trainer_data import trainer_pokemon_counts, trainer_win_counts
from stochastic_gradient_descent import minimize_stochastic
def squared_error(x_i, y_i, theta):
alpha, beta = theta
return error(alpha, beta, x_i, y_i)
def squared_error_gradient(x_i, y_i, theta):
alpha, beta, = theta
return [
-2 * error(alpha, beta, x_i, y_i), # alpha partial derivative
-2 * error(alpha, beta, x_i, y_i) * x_i # beta partial derivative
]
random.seed(0)
theta = [random.random(), random.random()]
alpha, beta = minimize_stochastic(squared_error, squared_error_gradient,
trainer_pokemon_counts, trainer_win_counts,
theta, 0.0000001)
# print('with minimize stochastic')
# print('trainer_alpha = %s, trainer_beta = %s' % (alpha, beta))
def estimate_beta(x, y):
beta_initial = [random.random() for x_i in x]
return minimize_stochastic(squared_error, squared_error_gradient, x, y,
beta_initial, 0.001)
