-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlda_svm.py
147 lines (125 loc) · 5.89 KB
/
lda_svm.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
import argparse, os, time, pickle
import tomotopy as tp
import numpy as np
import matplotlib.pyplot as plt
from sklearn.linear_model import SGDClassifier
from sklearn.svm import LinearSVC
from sklearn.metrics import accuracy_score
from data import *
from util import *
from lda_trial import *
def main(args):
trainset = IMDBDataset('train', data_limit=20_000)
validset = IMDBDataset('valid')
if args.task == 'k':
n = 20000
ks = [2, 5, 10, 20, 50]
train_accu = [0 for _ in range(len(ks))]
test_accu = [0 for _ in range(len(ks))]
times = [0 for _ in range(len(ks))]
for _ in range(3):
for idx, k in enumerate(ks):
print(f'{k}.')
start = time.time()
trained_model, final_metric = tp_one_trial(trainset, args.model, k, n,
3, 5, # args.burn_in,
max_iter=1000, stop_increase=5, metric='ll')
times[idx] += time.time() - start
lda_x, lda_y = load_LDA_data_batch(trained_model, trainset)
model = LinearSVC()
model.fit(lda_x, lda_y)
prediction = model.predict(lda_x)
ground_truth = lda_y
print(f'Train: {100*accuracy_score(prediction, ground_truth):6.5f}%')
train_accu[idx] += accuracy_score(prediction, ground_truth)
lda_x, lda_y = load_LDA_data_batch(trained_model, validset)
prediction = model.predict(lda_x)
ground_truth = lda_y
print(f'Test: {100*accuracy_score(prediction, ground_truth):6.5f}%')
test_accu[idx] += accuracy_score(prediction, ground_truth)
print('-' * 100)
if args.task == 'n':
k = 20
ns = [3_000, 5_000, 10_000, 15_000, 20_000]
train_accu = [0 for _ in range(len(ns))]
test_accu = [0 for _ in range(len(ns))]
times = [0 for _ in range(len(ns))]
for _ in range(3):
for idx, k in enumerate(ns):
print(f'{k}.')
start = time.time()
trained_model, final_metric = tp_one_trial(trainset, args.model, k, n,
3, 5, # args.burn_in,
max_iter=1000, stop_increase=5, metric='ll')
times[idx] += time.time() - start
lda_x, lda_y = load_LDA_data_batch(trained_model, trainset)
model = LinearSVC()
model.fit(lda_x, lda_y)
prediction = model.predict(lda_x)
ground_truth = lda_y
print(f'Train: {100*accuracy_score(prediction, ground_truth):6.5f}%')
train_accu[idx] += accuracy_score(prediction, ground_truth)
lda_x, lda_y = load_LDA_data_batch(trained_model, validset)
prediction = model.predict(lda_x)
ground_truth = lda_y
print(f'Test: {100*accuracy_score(prediction, ground_truth):6.5f}%')
test_accu[idx] += accuracy_score(prediction, ground_truth)
print('-' * 100)
train_accu = [accu / rep for accu in train_accu]
test_accu = [accu / rep for accu in test_accu]
times = [accu / rep for accu in times]
print(train_accu)
print(test_accu)
print(times)
print(task, args.model)
if __name__ == '__main__':
cur_path = os.path.abspath(os.path.dirname(__file__))
result_path = os.path.join(cur_path, 'result')
make_dirs(result_path)
parser = argparse.ArgumentParser(description="Stat700proj experiment")
parser.add_argument("--model", type=str, choices=['lda', 'ctm', "slda"], default='lda')
parser.add_argument("--rep_times", type=int, default=3)
# train
parser.add_argument("--task", type=str, default='k')
parser.add_argument("--n", type=int, default=20_000)
parser.add_argument("--k", type=int, default=20)
# Other
parser.add_argument("--rm_top", type=int, default=5)
parser.add_argument("--min_cf", type=int, default=3)
# parser.add_argument("--burn_in", type=int, default=500)
args = parser.parse_args()
trainset = IMDBDataset('train', data_limit=20_000)
validset = IMDBDataset('valid')
# n = 20000
# ks = [2, 5, 10, 20, 50]
# train_accu = [0 for _ in range(len(ks))]
# test_accu = [0 for _ in range(len(ks))]
# times = [0 for _ in range(len(ks))]
# for _ in range(3):
# for idx, k in enumerate(ks):
# print(f'{k}.')
# start = time.time()
# trained_model, final_metric = tp_one_trial(trainset, args.model, k, n,
# 3, 5, # args.burn_in,
# max_iter=1000, stop_increase=5, metric='ll')
# times[idx] += time.time() - start
# lda_x, lda_y = load_LDA_data_batch(trained_model, trainset)
# model = LinearSVC()
# model.fit(lda_x, lda_y)
# prediction = model.predict(lda_x)
# ground_truth = lda_y
# print(f'Train: {100*accuracy_score(prediction, ground_truth):6.5f}%')
# train_accu[idx] += accuracy_score(prediction, ground_truth)
# lda_x, lda_y = load_LDA_data_batch(trained_model, validset)
# prediction = model.predict(lda_x)
# ground_truth = lda_y
# print(f'Test: {100*accuracy_score(prediction, ground_truth):6.5f}%')
# test_accu[idx] += accuracy_score(prediction, ground_truth)
# print('-' * 100)
# train_accu = [accu / rep for accu in train_accu]
# test_accu = [accu / rep for accu in test_accu]
# times = [accu / rep for accu in times]
# print(train_accu)
# print(test_accu)
# print(times)
main(args)