from hw5.libs.common.dataset import Dataset
from hw5.libs.common.util import plot_digit_data
import matplotlib.pyplot as plt
def index_digit_ones(Y_train):
idx_ones = []
for i in range(len(Y_train)):
if int(Y_train[i]) == 1:
idx_ones.append(i)
return idx_ones
if __name__ == '__main__':
# dataset = Dataset(train_data=1000, test_data=100)
# dataset = Dataset(train_data=40, test_data=10)
dataset = Dataset(train_data=80, test_data=20)
# dataset = Dataset()
X_train, Y_train, X_test, Y_test = dataset.get_dataset()
print("before PCA = ", X_train.shape)
# Dimensional reduction
# pca = PCA(n_components=2, whiten=False)
# pca = PCA(n_components=64, whiten=False)
pca = PCA(n_components=80, whiten=False)
X_train = pca.fit_transform(X_train)
print("after PCA = ", X_train.shape)
# print(X_train[1].shape)
print(Y_train)
idx_ones = index_digit_ones(Y_train)
# https://deeplearningcourses.com/c/data-science-supervised-machine-learning-in-python
# https://www.udemy.com/data-science-supervised-machine-learning-in-python
# Code: https://github.com/lazyprogrammer/machine_learning_examples/blob/master/supervised_class/knn.py
import matplotlib.pyplot as plt
from hw5.libs.algo.knn import KNN
from hw5.libs.common.dataset import Dataset
from hw5.libs.common.util import int_to_tuple, save_to_csv
from datetime import datetime
if __name__ == '__main__':
# dataset = Dataset(train_data=80, test_data=20)
# dataset = Dataset(train_data=800, test_data=200)
dataset = Dataset()
X_train, Y_train, X_test, Y_test = dataset.get_dataset()
train_scores = []
test_scores = []
exec_times = []
best_test_score = 0
stored_accuracy = [0, 0] # [<acc>, <k-th>]
# Define number of iteration (K)
K = 50
ks = int_to_tuple(K) # used to plot the results
# Start KNN: Scratch
print("Evaluating Scratch KNN")
for i in range(K):
k = i+1
print("\nk =", k)
knn = KNN(k)
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
# import matplotlib.pyplot as plt
import seaborn as sns
sns.set() # for plot styling
import numpy as np
from hw6.libs.algo.kmeans import MyKMeans
if __name__ == '__main__':
# dataset = Dataset(train_data=1000, test_data=100)
# dataset = Dataset(train_data=20, test_data=10)
# dataset = Dataset(train_data=40, test_data=10)
# dataset = Dataset(train_data=80, test_data=20)
dataset = Dataset(train_data=800, test_data=200)
# dataset = Dataset()
# X_train, Y_train, X_test, Y_test = dataset.get_dataset()
init_X_train, init_Y_train, _, _ = dataset.get_dataset()
acc_scores = [] # ONLY collect the highest accuracy!
# Simulate clustering in K times.
K = 50
# K = 5
ks = int_to_tuple(K) # used to plot the results
# simulate different number of clusters
n_clusters = [2, 3, 4, 5, 6, 7, 8, 9, 10] # from n=2 ~ n=10 (max)
# n_clusters = [2, 3] # from n=2 ~ n=10 (max)
# Start simulation ...
# Define value of clustering simulation
# K = 9
# K = 1 # total number of simulations
K = 4 # total number of simulations
dataset_size = 200 # initialize started dataset size here
# dataset_step_size = 100 # e.g. datasets = {100, 200, 300, ..., 1000}
dataset_step_size = 200 # e.g. datasets = {100, 200, 300, ..., 1000}
ratio_train, ratio_test = 0.8, 0.2 # should be total value=1 ! e.g. ratio = { (80,20), (160,40), ..., (800,200) }
datasets = []
ds = [] # list of dataset size in each iterations
for i in range(K):
# print(" >> dataset_size = ", dataset_size)
train_data = int(dataset_size * ratio_train)
test_data = dataset_size - train_data
dataset = Dataset(train_data=train_data, test_data=test_data)
X_train, Y_train, X_test, Y_test = dataset.get_dataset()
datasets.append({
"X_train": X_train,
"Y_train": Y_train,
"X_test": X_test,
"Y_test": X_train
})
ds.append(dataset_size)
dataset_size += dataset_step_size
# print(" >> train_data, test_data = ", train_data, test_data)
# print(" >> ds = ", ds)
# Start KMeans: sklearn
print("Evaluating sklearn KMeans")
Based on the method used in "Learning To Remember Rare Events"
by Lukasz Kaiser, Ofir Nachun, Aurko Roy, and Samy Bengio
Paper: https://openreview.net/pdf?id=SJTQLdqlg
'''
import numpy as np
import theano
import theano.tensor as T
import time
import os
from hw5.libs.common.dataset import Dataset
from hw5.libs.common.util import int_to_tuple
if __name__ == '__main__':
dataset = Dataset(train_data=10, test_data=3)
# dataset = Dataset()
X_train, Y_train, X_test, Y_test = dataset.get_dataset()
train_scores = []
test_scores = []
print(" > Total X_train = ", len(X_train))
print(" > Total X_test = ", len(X_test))
print(">>> Y_test = ", Y_test)
# Define number of iteration (K)
K = 20
ks = int_to_tuple(K) # used to plot the results
def l2_normalize(x, dim, epsilon=1e-12):
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
# import matplotlib.pyplot as plt
import seaborn as sns
sns.set() # for plot styling
import numpy as np
from datetime import datetime
from hw6.libs.algo.gmm import MyGMM
if __name__ == '__main__':
# dataset = Dataset(train_data=1000, test_data=100)
# dataset = Dataset(train_data=20, test_data=10)
# dataset = Dataset(train_data=40, test_data=10)
# dataset = Dataset(train_data=80, test_data=20)
dataset = Dataset(train_data=800, test_data=200)
# dataset = Dataset()
X_train, Y_train, _, _ = dataset.get_dataset()
acc_scores = []
# Simulate clustering in K times.
K = 50
# K = 5
ks = int_to_tuple(K) # used to plot the results
# to make it easier to analyze, take only digit={0, 1, ...}
selected_digits = [1, 2, 3]
# selected_digits = [1, 2]
X_train, Y_train = filter_dataset(selected_digits, X_train, Y_train)
# Used for visualization only