感知機手寫

參考文獻：1.

2. 應用多元統計分析，高惠璇. 北京大學出版社

3. 統計學習方法第二版，李航

import numpy as np

class
model
:def
__init__
(self, w=[0
,0], b=
0, alpha=
0.2, max_counts=
5000):
self.w = np.array(w, dtype=np.float32)
self.b = b
self.l_rate = alpha
self.max_counts = max_counts
self.pred_list =
# self.data = data
defsign
(self, x, w, b)
:        y = np.dot(x, w)
+ b        return y
# 隨機梯度下降法
deffit
(self, x_train, y_train)
:        is_wrong =
false
count =
0while
not is_wrong:
count +=
1            wrong_count =
0for d in
range
(len
(x_train)):
x = x_train[d]
y = y_train[d]
if y * self.sign(x, self.w, self.b)
<=0:
self.w = self.w + self.l_rate * np.dot(y, x)
self.b = self.b + self.l_rate * y
wrong_count +=
1if self.max_counts == count:
print
("在設定的最大次數中，無法分開"
)break
if wrong_count ==0:
is_wrong =
true
return
'perceptron model!'
defpredict
(self, x_test, y_test)
:        self.y_test = y_test
for i in x_test:
if(np.dot(self.w, i)
+ self.b)
>0:
1)else:-
1)self.pred_array = np.array(np.array(self.pred_list)
== self.y_test)
return self.pred_array
defscore
(self, method=
"precision"):
if method ==
"precision"
:            pre_scr = np.array(self.pred_array, dtype=np.int8)
precision_scr = np.
sum(pre_scr)
/len
(pre_scr)
return precision_scr
elif method ==
"call"
:from collections import counter
class_dic = counter(self.y_test)
count =
0for i, j in
zip(self.pred_list, self.y_test)
:if i ==
1and j ==1:
count +=
1            call_scr = count / class_dic[1]
return call_scr
elif method ==
"f":
pre_scr = np.array(self.pred_array, dtype=np.int8)
precision_scr = np.
sum(pre_scr)
/len
(pre_scr)
from collections import counter
class_dic = counter(self.y_test)
count =
0for i, j in
zip(self.pred_list, self.y_test)
:if i ==
1and j ==1:
count +=
1            call_scr = count / class_dic[1]
return
(call_scr + precision_scr)
/2

機器學習 手寫感知機模型

資料集來自mnist資料集，主要利用numpy裡的matrix矩陣計算。演算法的實現主要在於對引數w和b的求解。演算法的推導過程參考李航 統計學習方法 推導最優化函式然後更新引數的過程。import numpy as np import pandas as pd import time defdat...

離線手寫中文簽名檢索 軟體設計方案

離線手寫中文簽名檢索 軟體設計方案 一 專案背景 我的工程實踐是離線 離線手寫中文簽名的智慧型檢索系統，首先通過人工智慧和深入學習技術，搭建乙個深度學習網路，再利用資料庫對網路進行訓練，最後通過自製框架整合到網頁端，搭建乙個簽名檢測系統，實現乙個在web端根據輸入的簽名從簽名資料庫中根據每個簽名的相...

Keras多層感知器識別手寫數字

from keras.utils import np utils import numpy as np np.random.seed 10 from keras.datasets import mnist x train image,y train label x test image,y test...