-
Notifications
You must be signed in to change notification settings - Fork 5
/
LinearReg.py
52 lines (43 loc) · 1.53 KB
/
LinearReg.py
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
class LinearReg():
def __init__(self,Method="Analytical"):
"""
Linear Regression with 2 Different Method
Param: Method -> "Analytical" or "GradientDes"
"""
self.Method = Method
def fit(self,x,y,lr=0.0001,epoch=10000,iniM=1.5,iniB=5):
"""
Prams:
x = X data
y = labels
lr = Learning rate
epoch = number of iteration
iniM = initial slope(m)
iniB = initial bias(b)
Param: Method -> "Analytical" or "GradientDes"
"""
#Analytical Solution of Linear Regression (with MSE loss/LSO )
if self.Method == "Analytical":
self.theta = (x.T @ y) / (x.T @ x)
#Gradient Descent Method
elif self.Method == "GradientDes":
n = x.shape[0]
self.m = iniM
self.b = iniB
for _ in range(epoch):
b_gradient = -2 * sum(y - self.m*x + self.b) / n
m_gradient = -2 * sum(x*(y - (self.m*x + self.b))) / n
self.b = self.b + (lr * b_gradient)
self.m = self.m - (lr * m_gradient)
def predict(self,xnew):
"""
Predict
Param:
xnew = list or 1-D array X data for prediction
"""
if self.Method == "Analytical":
newy = self.theta * xnew
return newy
elif self.Method == "GradientDes":
newy = self.m * xnew + self.b
return newy