- 邏輯回歸的推導過程：https://blog.csdn.net/ACM_hades/article/details/90448785
- 代碼主要實現了下面公式： W k + 1 = W k + λ X ( Y ? f W k ( X T ) ) W^{k+1}=W^k+λX(Y-f_{W^k } (X^T)) W k + 1 = W k + λ X ( Y ? f W k ? ( X T ) )
- 數據集 ：我們選擇MNIST數據集進行實驗，它包含各種手寫數字(0-9)圖片，圖片大小28*28。MNIST數據集本身有10個類別，為了將其變成二分類問題我們進行如下處理：label等于0的繼續等于0，label大于0改為1。這樣就將十分類的數據改為二分類的數據。
-
特征選擇
：可選擇的特征有很多，包括：
- 自己提取特征
- 將整個圖片作為特征向量
- HOG特征
- 我們將整個圖片作為特征(784=28×28)。
import
time
import
numpy
as
np
import
pandas
as
pd
from
sklearn
.
model_selection
import
train_test_split
from
sklearn
.
metrics
import
accuracy_score
class
Logistic
:
def
__init__
(
self
,
feature_len
)
:
self
.
weights
=
np
.
ones
(
(
feature_len
,
1
)
)
def
model_function
(
self
,
X
)
:
W_X
=
np
.
matmul
(
X
,
self
.
weights
)
temp_1
=
(
W_X
>=
0
)
.
astype
(
np
.
float
)
temp_0
=
(
W_X
<
0
)
.
astype
(
np
.
float
)
resut_1
=
1.0
/
(
1
+
np
.
exp
(
-
temp_1
*
W_X
)
)
*
temp_1
#W_X為負數是，因為參數值inx很大時，exp(inx)可能會發生溢出，所以修改計算方式
resut_0
=
np
.
exp
(
temp_0
*
W_X
)
/
(
1
+
np
.
exp
(
temp_0
*
W_X
)
)
*
temp_0
return
resut_1
+
resut_0
def
train
(
self
,
Data
,
label
)
:
#訓練
label
=
label
.
reshape
(
(
-
1
,
1
)
)
alpha
=
0.01
max_iter
=
500
for
i
in
range
(
max_iter
)
:
#迭代
pres
=
self
.
model_function
(
Data
)
error
=
label
-
pres
#預測值和標簽值所形成的誤差
self
.
weights
=
self
.
weights
+
alpha
*
np
.
matmul
(
Data
.
T
,
error
)
#權重的更新
def
predict
(
self
,
Data
)
:
return
self
.
model_function
(
Data
)
.
reshape
(
-
1
)
if
__name__
==
'__main__'
:
print
(
'Start read data'
)
S
=
time
.
time
(
)
raw_data
=
pd
.
read_csv
(
'./lihang_book_algorithm-master/data/train_binary.csv'
)
# 讀取數據
data
=
raw_data
.
values
# 獲取數據
print
(
"data shape:"
,
data
.
shape
)
imgs
=
data
[
:
,
1
:
]
labels
=
data
[
:
,
0
]
print
(
"imgs shape:"
,
imgs
.
shape
)
imgs
=
np
.
concatenate
(
(
imgs
,
np
.
ones
(
(
imgs
.
shape
[
0
]
,
1
)
)
)
,
axis
=
1
)
#拼接常數項
print
(
"imgs shape:"
,
imgs
.
shape
)
print
(
"labels shape:"
,
labels
.
shape
)
print
(
"label:"
,
list
(
set
(
labels
)
)
)
Model
=
Logistic
(
imgs
.
shape
[
-
1
]
)
# 選取 2/3 數據作為訓練集， 1/3 數據作為測試集
train_features
,
test_features
,
train_labels
,
test_labels
=
train_test_split
(
imgs
,
labels
,
test_size
=
0.33
,
random_state
=
23323
)
print
(
"train data count ：%d"
%
len
(
train_labels
)
)
print
(
"test data count ：%d"
%
len
(
test_labels
)
)
print
(
'read data cost '
,
time
.
time
(
)
-
S
,
' second'
)
print
(
'Start training'
)
S
=
time
.
time
(
)
Model
.
train
(
train_features
,
train_labels
)
print
(
'training cost '
,
time
.
time
(
)
-
S
,
' second'
)
print
(
'Start predicting'
)
S
=
time
.
time
(
)
test_predict
=
Model
.
predict
(
test_features
)
print
(
'predicting cost '
,
time
.
time
(
)
-
S
,
' second'
)
score
=
accuracy_score
(
test_labels
,
test_predict
)
print
(
"The accruacy socre is "
,
score
)
結果：
Start read data
data shape: (42000, 785)
imgs_origin shape: (42000, 784)
imgs shape: (42000, 785)
labels shape: (42000,)
label: [0, 1]
train data count ：28140
test data count ：13860
read data cost 4.148890018463135 second
Start training
training cost 15.161401748657227 second
Start predicting
predicting cost 0.007978200912475586 second
The accruacy socre is 0.9892496392496393
更多文章、技術交流、商務合作、聯系博主
微信掃碼或搜索：z360901061
微信掃一掃加我為好友
QQ號聯系： 360901061
您的支持是博主寫作最大的動力，如果您喜歡我的文章，感覺我的文章對您有幫助，請用微信掃描下面二維碼支持博主2元、5元、10元、20元等您想捐的金額吧，狠狠點擊下面給點支持吧，站長非常感激您！手機微信長按不能支付解決辦法：請將微信支付二維碼保存到相冊，切換到微信，然后點擊微信右上角掃一掃功能，選擇支付二維碼完成支付。
【本文對您有幫助就好】元
