def load_cv_data(validation_data_folder, percent):
"""
custom cv data
"""
all_folds = []
for fold in [1, 2, 3, 4, 5]:
X_file = f"{validation_data_folder}/Fold{fold}/X_{percent}%.txt"
y_file = f"{validation_data_folder}/Fold{fold}/y_{percent}%.txt"
X = utils.load_data_from_txt_file(X_file)
y = utils.load_data_from_txt_file(y_file, True)
all_folds.append((X, y))
return all_folds
def load_learning_curve_data(learning_curve_data_folder):
"""
Load all learning curve data
Args:
learning_curve_data_folder (str):
Directory to the folder containing the data. This
folder must contain the following files
X_train_10%.txt, X_train_20%.txt,..., X_train_100%.txt
y_train_10%.txt, y_train_10%.txt,..., y_train_100%.txt
Returns:
(all_X_train, all_y_train)
all_X_train is a list of 10 numpy arrays, with increasing
number of rows but with the same number of columns (features)
all_y_train is a list of 10 numpy arrays, with increasing number
of elements
Example usage
>>> subsets_X, subsets_y = load_learning_curve_data("/path/to/folder/with/data")
>>> for i, X in enumerate(subsets_X):
>>> y = subsets_y[i]
>>> # Train on X and y
"""
all_X_train = []
for percent in range(10, 101, 10): # For percent from 10, 20, ..., 100
X_file = f"{learning_curve_data_folder}/X_train_{percent}%.txt"
X = utils.load_data_from_txt_file(X_file)
all_X_train.append(np.asarray(X))
return np.asarray(all_X_train)
def load_data_custom(learning_curve_data_folder):
"""
I do what I want
"""
all_X_train = []
all_y_train = []
for percent in range(10, 101, 90): # For percent from 10, 20, ..., 100
X_file = f"{learning_curve_data_folder}/X_train_{percent}%.txt"
y_file = f"{learning_curve_data_folder}/y_train_{percent}%.txt"
X = utils.load_data_from_txt_file(X_file)
y = utils.load_data_from_txt_file(y_file, True)
all_X_train.append(X)
all_y_train.append(y)
return (all_X_train, all_y_train)
def generate_bmeans():
part_bmean = []
for num in range(1, 6): # For percent from 10, 20, ..., 100
X_file = f"P3/MeanInitialization/Part_b/mu_k_{num}.txt"
X = utils.load_data_from_txt_file(X_file)
part_bmean.append(np.asarray(X))
part_bmean = np.asarray(part_bmean)
return part_bmean
def load_all_cross_validation_data(validation_data_folder):
"""
Load all data to do cross validation experiment
Args:
validation_data_folder (str):
Directory to the folder containing the data
This directory must contains 5 sub-directories:
Fold1
Fold2
Fold3
Fold4
Fold5
Returns:
all_folds (list)
all_folds is a list of 5 elements. Each element is
a tuple (X,y) where
X is a numpy array of shape (N, d)
y is a numpy array of shape (N,)
Example usage:
>>> all_folds = load_all_cross_validation_data("/path/to/folder/with/CV-data")
>>> fold_number = 2 # Pick fold number 3 as leave out fold
>>> leave_out_data, training_data = partition_cross_validation_fold(all_folds, fold_number)
"""
all_folds = []
for fold in [1, 2, 3, 4, 5]:
X_file = f"{validation_data_folder}/X_train_fold{fold}.txt"
y_file = f"{validation_data_folder}/X_test_fold{fold}.txt"
X = utils.load_data_from_txt_file(X_file)
y = utils.load_data_from_txt_file(y_file)
all_folds.append((X, y))
return all_folds
# # Gaussian Mixture Models
import numpy as np
import matplotlib.pyplot as plt
import utils
from scipy.stats import multivariate_normal
from gmm import GaussianMixtureModel
X_test_all = utils.load_data_from_txt_file("P3/X_test.txt")
test_data = utils.load_data_from_txt_file("P3/X_test.txt")
def plot_contour_gaussian(ax, mean, covariance, eps=1e-2):
""" Plot the contour of a 2d Gaussian distribution with given mean and
covariance matrix
Args:
ax (matplotlib.axes.Axes):
Subplot used to plot the contour
mean (numpy.array):
Mean of the gaussian distribution
covariance (numpy.array):
Covariance matrix of the distribution
eps:
The cut off to draw the contour plot. The higher the value,
the smaller the contour plot.
Returns:
None
"""