import lad as lad
import random_dataset_utils as rdu
import plot_utils as pu
n_runs = 10000
n_samples = 500
p_outliers = 0.10
## Let's define some true values for the w vector, which
## later we will try to estimate from the noisy data.
## Note that w_true is a vertical vector
w_true = np.array([[1], \
[2]])
(X, Y_true) = rdu.generate_linear_relation_dataset(w_true, n_samples)
w_hat_list = [
] ## This is where we are going to store all the estimated w_hat's
w_hat_outliers_list = [] ## and this is for the case of outliers
runs_so_far = 0
while (runs_so_far < n_runs):
## Add gaussian noise to all points
Y = rdu.add_gaussian_noise(Y_true, 0, 1, 1)
## Add 1% of outliers (gaussian noise with much higher variance!)
Y_outliers = rdu.add_gaussian_noise(Y, 0, 5, p_outliers)
w_hat = lad.solve_lad_soft(X, Y, 1000) ## w_hat is a vertical vector
w_hat_list.append(w_hat)
w_hat_outliers = lad.solve_lad_soft(X, Y_outliers,
1000) ## w_hat is a vertical vector
import numpy as np
import sys
sys.path.append('../utils')
import lad as lad
import random_dataset_utils as rdu
n_x_dimensions = 1
n_y_dimensions = 1
n_runs_left = 10000
n_samples = 500
## Let's define some true values for the w vector, which
## later we will try to estimate from the noisy data.
linear_weigths = np.array([[1], [2]])
(X, Y) = rdu.generate_linear_relation_dataset(linear_weigths, n_samples)
## Add a moderate amount of gaussion noise to all points
Y = rdu.add_gaussian_noise(Y, 0, 1, 1)
w_hat = lad.solve_lad_soft(X, Y, 1000000)
print w_hat