-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_GA.py
374 lines (311 loc) · 17.1 KB
/
run_GA.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
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
# A genetic algorithm (GA) optimizing a set of miRNA-based cell classifiers for in situ cancer classification.
# Written by Melania Nowicka, FU Berlin, 2019.
import sys
import random
import time
import argparse
import numpy
import configparser
import genetic_algorithm
import preproc
import eval
numpy.random.seed(1)
random.seed(1)
# algorithm parameters read from the command line
def parse_parameters(args):
"""
Parses command line parameters.
Parameters
----------
args : list
list of command line arguments
"""
# parameter parser
parser = argparse.ArgumentParser(description='A genetic algorithm (GA) optimizing a set of miRNA-based cell '
'classifiers for in situ cancer classification. Written by Melania '
'Nowicka, FU Berlin, 2019.\n\n')
# adding arguments
parser.add_argument('--train', '--dataset-filename-train', dest="dataset_filename_train",
help='train data set file name')
parser.add_argument('--test', '--dataset-filename-test', dest="dataset_filename_test", default=None,
help='test data set file name')
parser.add_argument('--config', '--config', dest="config", type=str, default=None,
help='config file name')
parser.add_argument('--rules', '--rules', dest="rule_list", type=str, default=None,
help='list of pre-optimized rules')
parser.add_argument('--repeats', '--repeats', dest="repeats", type=int, default=1,
help='number of training repeats')
parser.add_argument('--filter', '--filter-data', dest="filter_data", type=str, default='t',
help='filter data')
parser.add_argument('--discretize', '--discretize-data', dest="discretize_data", type=str, default='t',
help='discretize data')
parser.add_argument('--mbin', '--m-bin', dest="m_bin", type=int, default=50,
help='m segments')
parser.add_argument('--abin', '--a-bin', dest="a_bin", type=float, default=0.5,
help='binarization alpha')
parser.add_argument('--lbin', '--l-bin', dest="l_bin", type=float, default=0.1,
help='binarization lambda')
parser.add_argument('-c', '--classifier-size', dest="classifier_size", type=int, default=5,
help='classifier size')
parser.add_argument('-a', '--evaluation-threshold', dest="evaluation_threshold", default=None,
help='evaluation threshold alpha')
parser.add_argument('-w', '--bacc-weight', dest="bacc_weight", default=0.5, type=float,
help='multi-objective function weight')
parser.add_argument('-u', '--uniqueness', dest="uniqueness", default='t', type=str,
help='uniqueness of inputs')
parser.add_argument('-i', '--iterations', dest="iterations", type=int, default=30,
help='number of iterations without improvement')
parser.add_argument('-f', '--fixed-iterations', dest="fixed_iterations", type=int, default=0,
help='fixed number of iterations')
parser.add_argument('-p', '--population-size', dest="population_size", type=int, default=300,
help='population size')
parser.add_argument('--elitism', '--elitism', dest="elitism", type=float, default=1,
help='copy fraction of current best solutions to the population')
parser.add_argument('--poptfrac', '--p-opt-frac', dest="p_opt_frac", type=float, default=0.5,
help='pre-optimized fraction of population')
parser.add_argument('-x', '--crossover-probability', dest="crossover_probability", default=0.8, type=float,
help='probability of crossover')
parser.add_argument('-m', '--mutation-probability', dest="mutation_probability", default=0.1, type=float,
help='probability of mutation')
parser.add_argument('-t', '--tournament-size', dest="tournament_size", default=0.2, type=float,
help='tournament size')
# parse arguments
params = parser.parse_args(args)
parameters = [params.dataset_filename_train, params.dataset_filename_test, params.rule_list, params.repeats,
params.filter_data, params.discretize_data, params.m_bin, params.a_bin, params.l_bin,
params.classifier_size, params.evaluation_threshold, params.bacc_weight,
params.uniqueness, params.iterations, params.fixed_iterations, params.population_size,
params.elitism, params.p_opt_frac, params.crossover_probability,
params.mutation_probability, params.tournament_size]
return parameters
# function to read parameters from config file
def read_from_config(train_datafile, test_datafile, rule_file, config_file_name):
config_file = configparser.ConfigParser()
config_file.read(config_file_name)
filter_data = config_file.getboolean("DATA PREPROC", "Filtering")
discretize_data = config_file.getboolean("DATA PREPROC", "Discretization")
m_bin = int(config_file['BINARIZATION PARAMETERS']['MSegments'])
a_bin = float(config_file['BINARIZATION PARAMETERS']['AlphaBin'])
l_bin = float(config_file['BINARIZATION PARAMETERS']['LambdaBin'])
classifier_size = int(config_file['CLASSIFIER PARAMETERS']['ClassifierSize'])
evaluation_threshold = float(config_file['CLASSIFIER PARAMETERS']['Alpha'])
bacc_weight = float(config_file['OBJECTIVE FUNCTION']['Weight'])
uniqueness = config_file.getboolean("OBJECTIVE FUNCTION", "Uniqueness")
iterations = int(config_file['GA PARAMETERS']['Iterations'])
fixed_iterations = int(config_file['GA PARAMETERS']['FixedIterations'])
population_size = int(config_file['GA PARAMETERS']['PopulationSize'])
elitism = config_file.getboolean("GA PARAMETERS", "Discretization")
popt_fraction = float(config_file['GA PARAMETERS']['PoptFraction'])
crossover_probability = float(config_file['GA PARAMETERS']['CrossoverProbability'])
mutation_probability = float(config_file['GA PARAMETERS']['MutationProbability'])
tournament_size = float(config_file['GA PARAMETERS']['TournamentSize'])
args = [train_datafile, test_datafile, rule_file, filter_data, discretize_data, m_bin, a_bin, l_bin,
classifier_size, evaluation_threshold, bacc_weight, uniqueness, iterations, fixed_iterations,
population_size, elitism, popt_fraction, crossover_probability, mutation_probability, tournament_size]
return args
# process parameters and data and run algorithm
def process_and_run(args):
"""
Processes data and parameters and runs the algorithm.
Parameters
----------
args : list
list of command line arguments
Returns
-------
train_bacc : float
training balanced accuracy
test_bacc : float
test balanced accuracy
updates : int
number of best score updates
training_time : float
training time
first_global : float
first global best score
first_avg_pop : float
first population average score
"""
train_datafile, test_datafile, rule_list, filter_data, discretize_data, m_bin, a_bin, l_bin, classifier_size, \
evaluation_threshold, bacc_weight, uniqueness, iterations, fixed_iterations, population_size, elitism, \
popt_fraction, crossover_probability, mutation_probability, tournament_size = args
print("##PARAMETERS##")
if filter_data == 't':
print("FILTERING: ", "on")
filter_data = True
else:
print("FILTERING: ", "off")
filter_data = False
if discretize_data == 't':
print("DISCRETIZE: ", "on")
print("DISCRETIZATION M: ", m_bin)
print("DISCRETIZATION ALPHA: ", a_bin)
print("DISCRETIZATION LAMBDA: ", l_bin)
else:
print("DISCRETIZE: ", "off")
print("EVALUATION THRESHOLD: ", evaluation_threshold)
print("MAX SIZE: ", classifier_size)
print("WEIGHT: ", bacc_weight)
if uniqueness == 't':
print("UNIQUENESS: ", "on")
uniqueness = True
else:
print("UNIQUENESS: ", "off")
uniqueness = False
if rule_list is not None:
print("POPULATION PRE-OPTIMIZATION: ", "on")
print("POPULATION PRE-OPTIMIZED FRACTION: ", popt_fraction)
print("GA PARAMETERS: ", "TC: ", iterations, ", PS: ", population_size, ", CP: ", crossover_probability, ", MP: ",
mutation_probability, ", TS: ", tournament_size)
print("\n##TRAIN DATA##")
# read the data
train_dataset, annotation, negatives, positives, features = preproc.read_data(train_datafile)
annotation = train_dataset["Annots"]
# discretize data
if discretize_data == 't':
print("\n##DISCRETIZATION##")
data_discretized, features, thresholds, feature_cdds = \
preproc.discretize_train_data(train_dataset, m_bin, a_bin, l_bin, True)
else:
data_discretized = train_dataset
feature_cdds = {}
bacc_weight = 1.0
print("\nTRAINING...")
start_train = time.time()
classifier, best_classifiers, updates, first_best_score, first_avg_pop = \
genetic_algorithm.run_genetic_algorithm(data_discretized, filter_data, iterations, fixed_iterations,
population_size, elitism, rule_list, popt_fraction, classifier_size,
evaluation_threshold, feature_cdds, crossover_probability,
mutation_probability, tournament_size, bacc_weight, uniqueness, True)
end_train = time.time()
training_time = end_train - start_train
print("TRAINING TIME: ", end_train - start_train)
# evaluate best classifier
classifier_score, train_bacc, errors, train_error_rates, train_additional_scores, cdd_score = \
eval.evaluate_classifier(classifier, data_discretized, annotation, positives, negatives, feature_cdds,
uniqueness, bacc_weight)
print("\n##TRAIN DATA SCORES##")
print("BACC: ", train_bacc)
print("CDD SCORE: ", cdd_score)
print("TPR: ", train_error_rates["tpr"])
print("TNR: ", train_error_rates["tnr"])
print("FNR: ", train_error_rates["fpr"])
print("FPR: ", train_error_rates["fnr"])
if test_datafile is not None:
print("\n##TEST DATA##")
# read test data
test_dataset, annotation, negatives, positives, features = preproc.read_data(test_datafile)
annotation = test_dataset["Annots"]
# discretize data
if discretize_data == 't':
print("\n##DISCRETIZATION##")
data_discretized = preproc.discretize_test_data(test_dataset, thresholds)
else:
data_discretized = test_dataset
feature_cdds = {}
bacc_weight = 1.0
# evaluate classifier
classifier_score, test_bacc, errors, test_error_rates, test_additional_scores, cdd_score = \
eval.evaluate_classifier(classifier, data_discretized, annotation, positives, negatives, feature_cdds,
uniqueness, bacc_weight)
print("\n##TEST DATA SCORES##")
print("BACC: ", test_bacc)
print("CDD SCORE: ", cdd_score)
print("TPR: ", test_error_rates["tpr"])
print("TNR: ", test_error_rates["tnr"])
print("FNR: ", test_error_rates["fpr"])
print("FPR: ", test_error_rates["fnr"])
else:
test_bacc = None
return train_bacc, test_bacc, updates, training_time, first_best_score, first_avg_pop
# repeat GA run
def repeat(args):
"""
Processes data and parameters and runs the algorithm.
Parameters
----------
args : list
list of command line arguments
"""
parameters = parse_parameters(args)
train_datafile = parameters[0]
test_datafile = parameters[1]
config = parameters[2]
rule_file = parameters[3]
if config is None:
# process parameters
[train_datafile, test_datafile, rule_file, repeats, filter_data, discretize_data, m_bin, a_bin, l_bin,
classifier_size, evaluation_threshold, bacc_weight, uniqueness, iterations, fixed_iterations, population_size,
elitism, popt_fraction, crossover_probability, mutation_probability, tournament_size] = parameters
args = [train_datafile, test_datafile, rule_file, filter_data, discretize_data, m_bin, a_bin, l_bin,
classifier_size, evaluation_threshold, bacc_weight, uniqueness, iterations, fixed_iterations,
population_size, elitism, popt_fraction, crossover_probability, mutation_probability, tournament_size]
else:
args = read_from_config(train_datafile, test_datafile, rule_file, config)
train_scores = []
test_scores = []
run_time = []
updates_list = []
first_scores = []
first_avg_population_scores = []
for i in range(0, repeats):
print("\nREPEAT ", i+1)
train_bacc, test_bacc, updates, train_time, first_best_score, first_avg_pop = process_and_run(args)
train_scores.append(train_bacc)
test_scores.append(test_bacc)
run_time.append(train_time)
updates_list.append(updates)
first_scores.append(first_best_score)
first_avg_population_scores.append(first_avg_pop)
print("\nRESULTS")
if repeats > 1:
print("AVG TRAIN: ", numpy.average(train_scores), " STDEV: ", numpy.std(train_scores, ddof=1))
if test_datafile is not None:
print("AVG TEST: ", numpy.average(test_scores), " STDEV: ", numpy.std(test_scores, ddof=1))
print("AVG UPDATES: ", numpy.average(updates_list), " STDEV: ", numpy.std(updates_list, ddof=1))
print("AVG TRAINING TIME: ", numpy.average(run_time), " STDEV: ", numpy.std(run_time, ddof=1))
print("AVG FIRST BEST SCORE: ", numpy.average(first_scores), " STDEV: ", numpy.std(first_scores, ddof=1))
print("AVG INITIAL POPULATION SCORE: ", numpy.average(first_avg_population_scores), " STDEV: ",
numpy.std(first_avg_population_scores, ddof=1))
if test_datafile is not None:
print("CSV;", numpy.average(train_scores), ";", numpy.std(train_scores, ddof=1), ";",
numpy.average(test_scores), ";", numpy.std(test_scores, ddof=1), ";",
numpy.average(updates_list), ";", numpy.std(updates_list, ddof=1), ";",
numpy.average(run_time), ";", numpy.std(run_time, ddof=1), ";",
numpy.average(first_scores), ";", numpy.std(first_scores, ddof=1), ";",
numpy.average(first_avg_population_scores), ";", numpy.std(first_avg_population_scores, ddof=1))
else:
print("CSV;", numpy.average(train_scores), ";", numpy.std(train_scores, ddof=1), ";",
numpy.average(updates_list), ";", numpy.std(updates_list, ddof=1), ";",
numpy.average(run_time), ";", numpy.std(run_time, ddof=1), ";",
numpy.average(first_scores), ";", numpy.std(first_scores, ddof=1), ";",
numpy.average(first_avg_population_scores), ";", numpy.std(first_avg_population_scores, ddof=1))
else:
print("AVG TRAIN: ", numpy.average(train_scores), " STDEV: ", 0.0)
if test_datafile is not None:
print("AVG TEST: ", numpy.average(test_scores), " STDEV: ", 0.0)
print("AVG UPDATES: ", numpy.average(updates_list), " STDEV: ", 0.0)
print("AVG TRAINING TIME: ", numpy.average(run_time), " STDEV: ", 0.0)
print("AVG FIRST BEST SCORE: ", numpy.average(first_scores), " STDEV: ", 0.0)
print("AVG INITIAL POPULATION SCORE: ", numpy.average(first_avg_population_scores), " STDEV: ",
0.0)
if test_datafile is not None:
print("CSV;", numpy.average(train_scores), ";", 0.0, ";",
numpy.average(test_scores), ";", 0.0, ";",
numpy.average(updates_list), ";", 0.0, ";",
numpy.average(run_time), ";", 0.0, ";",
numpy.average(first_scores), ";", 0.0, ";",
numpy.average(first_avg_population_scores), ";", 0.0)
else:
print("CSV;", numpy.average(train_scores), ";", 0.0, ";",
numpy.average(updates_list), ";", 0.0, ";",
numpy.average(run_time), ";", 0.0, ";",
numpy.average(first_scores), ";", 0.0, ";",
numpy.average(first_avg_population_scores), ";", 0.0)
if __name__ == "__main__":
start = time.time()
print('A genetic algorithm (GA) optimizing a set of miRNA-based distributed cell classifiers \n'
'for in situ cancer classification. Written by Melania Nowicka, FU Berlin, 2019.\n')
repeat(sys.argv[1:])
end = time.time()
print("TIME: ", end - start)