-
Notifications
You must be signed in to change notification settings - Fork 13
/
Copy pathTVSVM.py
230 lines (215 loc) · 10.5 KB
/
TVSVM.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
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
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
import numpy as np
from sklearn import preprocessing
from sklearn.base import BaseEstimator, ClassifierMixin
import KernelFunction as kf
import TwinPlane1
import TwinPlane2
#__copyright__ = ""
#__license__ = "GPL"
# __version__ = "1.1"
# __maintainer__ = "Arnav Kansal"
# __email__ = "[email protected]"
# __status__ = "Production"
class TwinSVMClassifier(BaseEstimator, ClassifierMixin):
def __init__(self,Epsilon1=0.1, Epsilon2=0.1, C1=1, C2=1,kernel_type=0,kernel_param=1,regulz1=1, regulz2=1,fuzzy=0,_estimator_type="classifier"):
self.Epsilon1=Epsilon1
self.Epsilon2=Epsilon2
self.C1=C1
self.C2=C2
self.regulz1 = regulz1
self.regulz2 = regulz2
self.fuzzy = fuzzy
self.kernel_type=kernel_type
self.kernel_param=kernel_param
def fit(self, X, Y):
assert (type(self.Epsilon1) in [float,int])
assert (type(self.Epsilon2) in [float,int])
assert (type(self.C1) in [float,int])
assert (type(self.C2) in [float,int])
assert (type(self.regulz1) in [float,int])
assert (type(self.regulz2) in [float,int])
assert (self.fuzzy in [0,1])
assert (type(self.kernel_param) in [float,int])
assert (self.kernel_type in [0,1,2,3])
####################fill in here
# mean -centering, doing std
#X_t = preprocessing.scale(X)
# Data Sorting, rearranging
Data = sorted(zip(Y,X), key=lambda pair: pair[0], reverse = True)
Total_Data = np.array([np.array(x) for y,x in Data])
A=np.array([np.array(x) for y,x in Data if (y==1)])
B=np.array([np.array(x) for y,x in Data if (y==0)])
# Radius, center of data calcs
if(self.fuzzy==1):
if(self.kernel_type==0):
rcenpos=0
rcenneg=0
xcenpos = np.true_divide(sum(A),len(A))
for a in A:
if(rcenpos<np.linalg.norm(a-xcenpos)):
rcenpos = np.linalg.norm(a-xcenpos)
xcenneg = np.true_divide(sum(B),len(B))
for b in B:
if(rcenneg<np.linalg.norm(b-xcenneg)):
rcenneg = np.linalg.norm(b-xcenneg)
self.xcenpos_ = xcenpos
self.xcenneg_ = xcenneg
self.rcenpos_ = rcenpos
self.rcenneg_ = rcenneg
else:
rcenpossq=-np.inf
termtemp1=0
for i in range(len(A)):
term1 = kf.kernelfunction(self.kernel_type,A[i],A[i],self.kernel_param)
term2 = 0
for j in range(len(A)):
term2 += kf.kernelfunction(self.kernel_type,A[j],A[i],self.kernel_param)
termtemp1 += kf.kernelfunction(self.kernel_type,A[i],A[j],self.kernel_param)
term2 = -2*term2/len(A)
rcenpossq = max(rcenpossq,term1+term2)
termtemp1 = termtemp1/(len(A)*len(A))
rcenpossq += termtemp1
rcennegsq=-np.inf
termtemp2=0
for i in range(len(B)):
term1 = kf.kernelfunction(self.kernel_type,B[i],B[i],self.kernel_param)
term2 = 0
for j in range(len(B)):
term2 += kf.kernelfunction(self.kernel_type,B[j],B[i],self.kernel_param)
termtemp2 += kf.kernelfunction(self.kernel_type,B[i],B[j],self.kernel_param)
term2 = -2*term2/len(B)
rcennegsq = max(rcennegsq,term1+term2)
termtemp2 = termtemp2/(len(B)*len(B))
rcennegsq += termtemp2
self.rcenpossq_ = rcenpossq
self.rcennegsq_ = rcennegsq
self.termtemp1_ = termtemp1
self.termtemp2_ = termtemp2
m1 = A.shape[0]
m2 = B.shape[0]
e1 = -np.ones((m1,1))
e2 = -np.ones((m2,1))
if(self.kernel_type==0): # no need to cal kernel
S = np.hstack((A,-e1))
R = np.hstack((B,-e2))
else:
S = np.zeros((A.shape[0],Total_Data.shape[0]))
for i in range(A.shape[0]):
for j in range(Total_Data.shape[0]):
S[i][j] = kf.kernelfunction(self.kernel_type,A[i],Total_Data[j],self.kernel_param)
S = np.hstack((S,-e1))
R = np.zeros((B.shape[0],Total_Data.shape[0]))
for i in range(B.shape[0]):
for j in range(Total_Data.shape[0]):
R[i][j] = kf.kernelfunction(self.kernel_type,B[i],Total_Data[j],self.kernel_param)
R = np.hstack((R,-e2))
#####################Calculation of Function Parameters(Equation of planes)
[w1,b1] = TwinPlane1.Twin_plane_1(R,S,self.C1,self.Epsilon1,self.regulz1)
[w2,b2] = TwinPlane2.Twin_plane_2(S,R,self.C2,self.Epsilon2,self.regulz2)
self.plane1_coeff_ = w1
self.plane1_offset_ = b1
self.plane2_coeff_ = w2
self.plane2_offset_ = b2
self.data_ = Total_Data
self.A_ = A
self.B_ = B
return self
def get_params(self, deep=True):
return {"Epsilon1": self.Epsilon1, "Epsilon2": self.Epsilon2, "C1": self.C1, "C2": self.C2, "regulz1": self.regulz1,
"regulz2":self.regulz2, "kernel_type": self.kernel_type, "kernel_param": self.kernel_param,"fuzzy": self.fuzzy}
def set_params(self, **parameters):
for parameter, value in parameters.items():
self.setattr(parameter, value)
return self
def predict(self, X, y=None):
#X_test = preprocessing.scale(X)
if(self.kernel_type==0): # no need to cal kernel
S = X
w1mod = np.linalg.norm(self.plane1_coeff_)
w2mod = np.linalg.norm(self.plane2_coeff_)
else:
S = np.zeros((self.data_.shape[0],self.data_.shape[0]))
for i in range(self.data_.shape[0]):
for j in range(self.data_.shape[0]):
S[i][j] = kf.kernelfunction(self.kernel_type,self.data_[i],self.data_[j],self.kernel_param)
w1mod = np.sqrt(np.dot(np.dot(self.plane1_coeff_.T,S),self.plane1_coeff_))
w2mod = np.sqrt(np.dot(np.dot(self.plane2_coeff_.T,S),self.plane2_coeff_))
S = np.zeros((X.shape[0],self.data_.shape[0]))
for i in range(X.shape[0]):
for j in range(self.data_.shape[0]):
S[i][j] = kf.kernelfunction(self.kernel_type,X[i],self.data_[j],self.kernel_param)
y1 = np.dot(S,self.plane1_coeff_)+ ((self.plane1_offset_)*(np.ones((X.shape[0],1))))
y2 = np.dot(S,self.plane2_coeff_)+ ((self.plane2_offset_)*(np.ones((X.shape[0],1))))
###############Compute test data predictions
yPredicted=np.zeros((X.shape[0],1))
distFromPlane1 = y1/w1mod #abs(np.dot(Z.transpose(),self.plane1_coeff_)+self.plane1_offset_)
distFromPlane2 = y2/w2mod #abs(np.dot(Z.transpose(),self.plane2_coeff_)+self.plane2_offset_)
for i in range(len(distFromPlane1)):
if (distFromPlane1[i]<distFromPlane2[i]):
yPredicted[i][0]=0;
else:
yPredicted[i][0]=1;
return yPredicted.transpose()[0]
def decision_function(self,X):
#X_test = preprocessing.scale(X)
# membership function:
# 1-(x+ - xi)/r+
# 1-(x_ - xi)/r_
#kernel
# 1-_/(|di^2|/(r+^2))
# 1-_/(|di^2|/(r-^2))
#fuzzy=0
if(self.fuzzy==1):
s1=[]
s2=[]
if(self.kernel_type==0): # no need to cal kernel
for i in range(len(X)):
s1.append(1-(np.linalg.norm(self.xcenpos_-X[i])/self.rcenpos_))
s2.append(1-(np.linalg.norm(self.xcenneg_-X[i])/self.rcenneg_))
else:
for i in range(len(X)):
dsquaredpos = kf.kernelfunction(self.kernel_type,X[i],X[i],self.kernel_param)
term1 = 0
for j in range(len(self.A_)):
term1 += kf.kernelfunction(self.kernel_type,self.A_[j],X[i],self.kernel_param)
term1 = -2*term1/len(self.A_)
dsquaredpos += term1
dsquaredpos += self.termtemp1_
s1.append(1-np.sqrt(dsquaredpos/self.rcenpossq_))
dsquaredneg = kf.kernelfunction(self.kernel_type,X[i],X[i],self.kernel_param)
term1 = 0
for j in range(len(self.B_)):
term1 += kf.kernelfunction(self.kernel_type,self.B_[j],X[i],self.kernel_param)
term1 = -2*term1/len(self.B_)
dsquaredneg += term1
dsquaredneg += self.termtemp2_
s2.append(1-np.sqrt(dsquaredneg/self.rcennegsq_))
s1 = np.array(s1)
s2 = np.array(s2)
return np.true_divide(s1,s1+s2)-0.5
else:
if(self.kernel_type==0): # no need to cal kernel
S = X
w1mod = np.linalg.norm(self.plane1_coeff_)
w2mod = np.linalg.norm(self.plane2_coeff_)
else:
S = np.zeros((self.data_.shape[0],self.data_.shape[0]))
for i in range(self.data_.shape[0]):
for j in range(self.data_.shape[0]):
S[i][j] = kf.kernelfunction(self.kernel_type,self.data_[i],self.data_[j],self.kernel_param)
w1mod = np.sqrt(np.dot(np.dot(self.plane1_coeff_.T,S),self.plane1_coeff_))
w2mod = np.sqrt(np.dot(np.dot(self.plane2_coeff_.T,S),self.plane2_coeff_))
S = np.zeros((X.shape[0],self.data_.shape[0]))
for i in range(X.shape[0]):
for j in range(self.data_.shape[0]):
S[i][j] = kf.kernelfunction(self.kernel_type,X[i],self.data_[j],self.kernel_param)
y1 = np.dot(S,self.plane1_coeff_)+ ((self.plane1_offset_)*(np.ones((X.shape[0],1))))
y2 = np.dot(S,self.plane2_coeff_)+ ((self.plane2_offset_)*(np.ones((X.shape[0],1))))
###############Compute test data predictions
yPredicted=np.zeros((X.shape[0],1))
distFromPlane1 = y1/w1mod
distFromPlane2 = y2/w2mod
###############Compute test data predictions
for i in range(len(distFromPlane1)):
yPredicted[i][0] = distFromPlane2[i]/(distFromPlane1[i]+distFromPlane2[i])-0.5
return yPredicted.transpose()[0]