-
Notifications
You must be signed in to change notification settings - Fork 0
/
KNN
96 lines (79 loc) · 3.05 KB
/
KNN
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
import numpy as np
import os
from PIL import Image
def load_data(path, index):
path = path + str(index)
counter = 0
images = np.array([[]])
for root,dirs,files in os.walk(path):
for f in files:
abs_path = os.path.join(root,f)
image = Image.open(abs_path)
image_array = np.array(image)
# flatten to 1-D vector; image_f: 784
image_f = image_array.flatten()
# add the new flattened image to our images set for training
if counter == 0:
images = np.concatenate((images, np.array([image_f])),axis=1)
else:
images = np.concatenate((images, np.array([image_f])))
# images' shape: (num_img, 784) # (1, num_img*784)
counter += 1
print(images.shape)
return images, counter # images: [num_image, 784]
def max_label(array):
u, counter = np.unique(array, return_counts=True)
index = np.argmax(counter)
return u[index]
def KNN(test_file_name, k):
train_images = np.array([[]])
train_label = np.array([])
for i in range(0,10):
label = str(i)
images, counter = load_data("D:\\HW2\\dataset\\train\\",i)
if i == 0:
train_images = images
else:
train_images = np.concatenate((train_images, images))
# train_images' shape: (total_train_img, 784)
for j in range(0, counter):
train_label = np.append(train_label, i)
print("label %d has been loaded"%(i))
test_images = np.array([[]])
test_label = np.array([])
for i in range(0,10):
label = str(i)
images, counter = load_data(test_file_name,i)
if i == 0:
test_images = images
else:
test_images = np.concatenate((test_images, images))
# test_images' shape: (total_test_img, 784)
for j in range(0, counter):
test_label = np.append(test_label, i)
print("Data loaded successfully")
print(train_images.shape) # output: (60000, 784)
print(test_images.shape) # output: (10000, 784)
counter = 0
error_times = 0
for test_image in test_images: # test_label's shape: [total_test_img, 784]
# test_image's shape: [784]
# E distance between this test_image and all train images
distance = np.array([])
for train_image in train_images:
d = np.linalg.norm(test_image - train_image)
distance = np.append(distance, d)
sort_index = distance.argsort()
top_k = np.array([])
for m in range(0, k):
top_k = np.append(top_k, train_label[sort_index[m]])
predict_label = max_label(top_k)
print(predict_label)
if predict_label != test_label[counter]:
error_times += 1
counter += 1
print(str(counter)+" test(s) completed!")
error_rate = error_times/counter
return error_rate
error_rate = KNN("D:\\HW2\\dataset\\test\\",10)
print("The error rate is %f"%(error_rate))