-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathsvm_generate_predictions.py
222 lines (178 loc) · 7.63 KB
/
svm_generate_predictions.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
##### IMPORTS #####
import numpy as np
import pandas as pd
import pickle
from tqdm import tqdm
from kernels import LinearKernel, GaussianKernel, PolynomialKernel, SpectrumKernel, MismatchKernel, SumKernel
from utils import create_kmer_set,m_neighbours,get_neighbours,load_neighbors,load_or_compute_neighbors
from classifiers.svm import SVM
##### PARAMETERS #####
kernel = 'sum' #'linear' 'rbf', 'poly', 'spectrum', 'mismatch' or 'sum'
C = 5.0 #Parameter C for SVM
gamma = 10.0 #Parameter gamma for SVM (only for 'rbf' or 'poly')
coef0 = 1.0 #Parameter coef0 for SVM (only for 'poly')
degree = 3 #Parameter degree for SVM (only for 'poly')
k = 12 #Parameter k for SVM (only for 'spectrum' and 'mismatch')
m = 2 #Parameter m for SVM (only for 'mismatch')
list_k = [5,8,10,12,13,15] #List of parameters k for sum of mismatch kernels (only for 'sum')
list_m = [1,1,1,2,2,3] #List of parameters m for sum of mismatch kernels (only for 'sum')
weights = [1.0,1.0,1.0,1.0,1.0,1.0] #List of weights for sum of mismatch kernels (only for 'sum')
shuffle = True #Shuffle the data
##### LOAD DATA #####
# shape (2000,1): string
X0_train = pd.read_csv("data/Xtr0.csv", sep=",", index_col=0).values
X1_train = pd.read_csv("data/Xtr1.csv", sep=",", index_col=0).values
X2_train = pd.read_csv("data/Xtr2.csv", sep=",", index_col=0).values
# shape (2000,100): float
X0_mat100_train = pd.read_csv("data/Xtr0_mat100.csv", sep=" ", header=None).values
X1_mat100_train = pd.read_csv("data/Xtr1_mat100.csv", sep=" ", header=None).values
X2_mat100_train = pd.read_csv("data/Xtr2_mat100.csv", sep=" ", header=None).values
# shape (2000,1): string
X0_test = pd.read_csv("data/Xte0.csv", sep=",", index_col=0).values
X1_test = pd.read_csv("data/Xte1.csv", sep=",", index_col=0).values
X2_test = pd.read_csv("data/Xte2.csv", sep=",", index_col=0).values
# shape (2000,100): float
X0_mat100_test = pd.read_csv("data/Xte0_mat100.csv", sep=" ", header=None).values
X1_mat100_test = pd.read_csv("data/Xte1_mat100.csv", sep=" ", header=None).values
X2_mat100_test = pd.read_csv("data/Xte2_mat100.csv", sep=" ", header=None).values
# shape (2000,1): 0 or 1
Y0_train = pd.read_csv("data/Ytr0.csv", sep=",", index_col=0).values
Y1_train = pd.read_csv("data/Ytr1.csv", sep=",", index_col=0).values
Y2_train = pd.read_csv("data/Ytr2.csv", sep=",", index_col=0).values
##### PREPROCESS DATA #####
#Rescaling labels
Y0_train = np.where(Y0_train == 0, -1, 1)
Y1_train = np.where(Y1_train == 0, -1, 1)
Y2_train = np.where(Y2_train == 0, -1, 1)
#If the kernel is 'mismatch', compute kmers neighbors
if kernel=='mismatch':
neighbours_0, kmer_set_0 = load_or_compute_neighbors(0,k,m)
neighbours_1, kmer_set_1 = load_or_compute_neighbors(1,k,m)
neighbours_2, kmer_set_2 = load_or_compute_neighbors(2,k,m)
#Some verifications for sum kernel
if kernel == 'sum':
assert(len(list_k)==len(list_m))
assert(len(weights)==len(list_m))
#Shuffling
if shuffle:
shuffling_0 = np.random.permutation(len(X0_mat100_train))
X0_train = X0_train[shuffling_0][:,0]
X0_mat100_train = X0_mat100_train[shuffling_0]
Y0_train = Y0_train[shuffling_0]
shuffling_1 = np.random.permutation(len(X1_mat100_train))
X1_train = X1_train[shuffling_1][:,0]
X1_mat100_train = X1_mat100_train[shuffling_1]
Y1_train = Y1_train[shuffling_1]
shuffling_2 = np.random.permutation(len(X2_mat100_train))
X2_train = X2_train[shuffling_2][:,0]
X2_mat100_train = X2_mat100_train[shuffling_2]
Y2_train = Y2_train[shuffling_2]
else:
X0_train = X0_train[:,0]
X1_train = X1_train[:,0]
X2_train = X2_train[:,0]
#Check if the kernel applies on matrices or strings
kernel_on_matrices = (kernel=='linear' or kernel=='rbf' or kernel=='poly')
#Put test matrices into the right format
X0_test = X0_test[:,0]
X1_test = X1_test[:,0]
X2_test = X2_test[:,0]
## PRINT CONFIGURATION ##
print("Kernel:", kernel)
print("C:", C)
if kernel == 'rbf' or kernel == 'poly':
print("Gamma:", gamma)
if kernel == 'poly':
print("Coef0:", coef0)
print("Degree:", degree)
if kernel== 'spectrum':
print("K:",k)
if kernel == 'sum':
print("List of Ks:",list_k)
print("List of Ms:",list_m)
print("Weights:", weights)
print()
##### APPLY SVM ON DATASET 0 #####
print("Applying SVM on dataset 0...")
if kernel=='linear':
svm = SVM(kernel=LinearKernel(),C=C)
elif kernel=='rbf':
svm = SVM(kernel=GaussianKernel(sigma=np.sqrt(0.5/gamma),normalize=False),C=C)
elif kernel=='poly':
svm = SVM(kernel=PolynomialKernel(gamma=gamma,coef0=coef0,degree=degree),C=C)
elif kernel=='spectrum':
svm = SVM(kernel=SpectrumKernel(k=k),C=C)
elif kernel=='mismatch':
svm = SVM(kernel=MismatchKernel(k=k, m=m, neighbours=neighbours_0, kmer_set=kmer_set_0,normalize=True), C=C)
elif kernel=='sum':
dataset_nbr = 0
kernels = []
for k,m in zip(list_k,list_m):
neighbours, kmer_set = load_or_compute_neighbors(dataset_nbr, k, m)
kernels.append(MismatchKernel(k=k, m=m, neighbours=neighbours, kmer_set=kmer_set, normalize = True))
svm = SVM(kernel=SumKernel(kernels=kernels, weights=weights), C=C)
if kernel_on_matrices:
svm.fit(X0_mat100_train, Y0_train)
pred_0 = svm.predict_classes(X0_mat100_test)
else:
svm.fit(X0_train, Y0_train)
pred_0 = svm.predict_classes(X0_test)
##### APPLY SVM ON DATASET 1 #####
print("Applying SVM on dataset 1...")
if kernel=='linear':
svm = SVM(kernel=LinearKernel(),C=C)
elif kernel=='rbf':
svm = SVM(kernel=GaussianKernel(sigma=np.sqrt(0.5/gamma),normalize=False),C=C)
elif kernel=='poly':
svm = SVM(kernel=PolynomialKernel(gamma=gamma,coef0=coef0,degree=degree),C=C)
elif kernel=='spectrum':
svm = SVM(kernel=SpectrumKernel(k=k),C=C)
elif kernel=='mismatch':
svm = SVM(kernel=MismatchKernel(k=k, m=m, neighbours=neighbours_1, kmer_set=kmer_set_1,normalize=True), C=C)
elif kernel=='sum':
dataset_nbr = 1
kernels = []
for k,m in zip(list_k,list_m):
neighbours, kmer_set = load_or_compute_neighbors(dataset_nbr, k, m)
kernels.append(MismatchKernel(k=k, m=m, neighbours=neighbours, kmer_set=kmer_set, normalize = True))
svm = SVM(kernel=SumKernel(kernels=kernels, weights=weights), C=C)
if kernel_on_matrices:
svm.fit(X1_mat100_train, Y1_train)
pred_1 = svm.predict_classes(X1_mat100_test)
else:
svm.fit(X1_train, Y1_train)
pred_1 = svm.predict_classes(X1_test)
##### APPLY SVM ON DATASET 2 #####
print("Applying SVM on dataset 2...")
if kernel=='linear':
svm = SVM(kernel=LinearKernel(),C=C)
elif kernel=='rbf':
svm = SVM(kernel=GaussianKernel(sigma=np.sqrt(0.5/gamma),normalize=False),C=C)
elif kernel=='poly':
svm = SVM(kernel=PolynomialKernel(gamma=gamma,coef0=coef0,degree=degree),C=C)
elif kernel=='spectrum':
svm = SVM(kernel=SpectrumKernel(k=k),C=C)
elif kernel=='mismatch':
svm = SVM(kernel=MismatchKernel(k=k, m=m, neighbours=neighbours_2, kmer_set=kmer_set_2,normalize=True), C=C)
elif kernel=='sum':
dataset_nbr = 2
kernels = []
for k,m in zip(list_k,list_m):
neighbours, kmer_set = load_or_compute_neighbors(dataset_nbr, k, m)
kernels.append(MismatchKernel(k=k, m=m, neighbours=neighbours, kmer_set=kmer_set, normalize = True))
svm = SVM(kernel=SumKernel(kernels=kernels, weights=weights), C=C)
if kernel_on_matrices:
svm.fit(X2_mat100_train, Y2_train)
pred_2 = svm.predict_classes(X2_mat100_test)
else:
svm.fit(X2_train, Y2_train)
pred_2 = svm.predict_classes(X2_test)
##### CREATE SUBMISSION FILE #####
pred = np.concatenate([pred_0.squeeze(),pred_1.squeeze(),pred_2.squeeze()])
pred = np.where(pred == -1, 0, 1)
pred_df = pd.DataFrame()
print(pred.shape)
print(pred)
pred_df['Bound'] = pred
pred_df.index.name = 'Id'
pred_df.to_csv('Yte.csv', sep=',', header=True)