-
Notifications
You must be signed in to change notification settings - Fork 1
/
syntactic_experiments.py
549 lines (455 loc) · 18.3 KB
/
syntactic_experiments.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
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
# Import tools
from type_system import *
from program import *
from cfg import *
from pcfg import *
from dsl import *
# Import DSL
from DSL.deepcoder import *
# Import algorithms
from Algorithms.heap_search import heap_search
from Algorithms.a_star import a_star
from Algorithms.threshold_search import threshold_search
from Algorithms.dfs import dfs
from Algorithms.bfs import bfs
from Algorithms.sqrt_sampling import sqrt_sampling
import pickle
import time
import random
import matplotlib.pyplot as plt
from math import log10
seed = 100
random.seed(seed)
deepcoder = DSL(semantics, primitive_types, no_repetitions)
t = Arrow(List(INT),List(INT))
deepcoder_PCFG_t = deepcoder.DSL_to_Random_PCFG(t, alpha = .7)
# first experiment: x = time, y = cumulative proba
# second experiment: x = number of programs enumerated, y = time
# third experiment: x = proba program, y = time/average time to find the program
# Set of algorithms where we need to reconstruct the programs
reconstruct = {dfs, bfs, threshold_search, a_star}
def create_dataset(PCFG):
'''
Create a dataset, which is a list of number_samples programs with proba in [1O^(-(i+1),1O^(-i)] for i in [imin, imax]
'''
dataset = []
size_dataset = [0 for _ in range(imax)]
finished = False
# gen = PCFG.sampling()
gen = sqrt_sampling(PCFG) # better than a simple sampling to get interesting programs
while(not finished):
program = next(gen) # format: a program
proba = PCFG.probability_program(PCFG.start, program)
i = int(-log10(proba))
if (i >= imin and i < imax and size_dataset[i] < number_samples):
# print("This program has probability %0.10f, it goes in bucket %u.\n" % (proba, i))
dataset.append((program,proba))
size_dataset[i] += 1
if size_dataset[i] == number_samples:
j = imin
finished = True
while(finished and j < imax):
finished = size_dataset[j] == number_samples
j += 1
# We sort the dataset by decreasing probability
dataset = sorted(dataset, key = lambda pair: pair[1], reverse = True)
# print(dataset)
return dataset
def run_algorithm(dsl, PCFG, algorithm, param):
'''
Run the algorithm until either timeout or 3M programs, and for each program record probability and time of output
'''
print("Running: %s" % algorithm.__name__)
# result = []
# seen = set()
result = {} # str(prog) : N, chrono, proba
N = 0
chrono = 0
gen = algorithm(PCFG, **param)
while (chrono < timeout and N < total_number_programs):
chrono -= time.perf_counter()
program = next(gen)
# if algorithm.__name__ == 'dfs':
# print(program)
chrono += time.perf_counter()
if algorithm.__name__ == 'bfs':
print(N)
if algorithm in reconstruct:
program = dsl.reconstruct_from_compressed(program, PCFG.start[0])
# if algorithm.__name__ == 'dfs':
# print(program)
# if N <= 10:
# print(algorithm.__name__, program)
hash_program = str(program)
if hash_program not in result:
N += 1
result[hash_program] = N, chrono, PCFG.probability_program(PCFG.start, program)
# result.append((program, PCFG.proba_term(PCFG.start, program), chrono))
print("Run successful, output %u programs" % len(result))
# print(result)
return result
def experiment_enumeration_time(result):
chrono_result = [chrono for (program, probability, chrono) in result]
return(chrono_result)
####################
# First experiment #
####################
# cumulative probability versus time
# parameters
timeout = 30 # in seconds
random.seed(seed)
total_number_programs = 150_000
dsl = deepcoder
pcfg = deepcoder_PCFG_t
threshold_probability = 1 # do not plot if cumulative proba greater than this threshold, otherwise hard to interpret
#title = "Cumulative probability versus time"
title = ""
recompute_from_scratch = False
#list_algorithms = [(heap_search, 'heap search', {'dsl' : dsl, 'environments': {}}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (sqrt_sampling, 'SQRT', {}), (a_star, 'A*', {})]
list_algorithms = [(heap_search, 'heap search', {}), (sqrt_sampling, 'SQRT', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (a_star, 'A*', {}), (bfs, 'bfs', {'beam_width' : 50000})]
if recompute_from_scratch:
for algo, algo_name, param in list_algorithms:
run_algo = run_algorithm(dsl, pcfg, algo, param)
with open("results_syntactic/run_1_" + algo_name + '_' + str(seed) + ".pickle","wb" ) as f:
pickle.dump(run_algo, f)
def experiment_cumulative_vs_time(run_algo):
result = [(run_algo[e][0], run_algo[e][1], run_algo[e][2]) for e in run_algo] #N, chrono, proba
result.sort(key = lambda x: x[0])
cumulative = 0
for i in range(0, len(result)):
proba = result[i][2]
cumulative+=proba
result[i] = (cumulative, result[i][1])
# result.append((cumulative, chrono))
return result
def plot_cumulative_vs_time(PCFG, list_algorithms):
'''Retrieve the results and plot'''
for i, (algorithm, algo_name, param) in enumerate(list_algorithms):
try:
f = open("results_syntactic/run_1_" + algo_name + '_' + str(seed) + ".pickle","rb")
run_algo = pickle.load(f)
f.close()
except:
print("algorithm not run yet")
assert(False)
processed_run_algo = experiment_cumulative_vs_time(run_algo)
plt.scatter([x for (x,y) in processed_run_algo], [y for (x,y) in processed_run_algo], label = algo_name, s = 8)
#plt.xlim((0,threshold_probability))
plt.legend()
plt.xlabel("cumulative probability")
plt.ylabel("time (in seconds)")
plt.title(title)
# plt.xscale('log')
plt.yscale('log')
#plt.show()
plt.savefig("results_syntactic/cumulative_time_%s.png" % seed, dpi=500, bbox_inches='tight')
plt.clf()
plot_cumulative_vs_time(pcfg, list_algorithms)
# #####################
# # Second experiment #
# #####################
# # Enumeration time: heap search versus A*
# # parameters
# # title = "Heap search versus A*"
# title = ""
# timeout = 50 # in seconds
# random.seed(seed)
# total_number_programs = 300_000
# dsl = deepcoder
# pcfg = deepcoder_PCFG_t
# list_algorithms = [(heap_search, 'heap search', {}), (a_star, 'A*', {})]
# recompute_from_scratch = False
# if recompute_from_scratch:
# for algo, algo_name, param in list_algorithms:
# run_algo = run_algorithm(dsl, pcfg, algo, param)
# with open("results_syntactic/run_2_" + algo_name + '_' + str(seed) + ".pickle","wb" ) as f:
# pickle.dump(run_algo, f)
# def experiment_enumeration_time(run_algo):
# result = [run_algo[e][1] for e in run_algo] #N, chrono, proba
# result.sort()
# return result
# def plot_enumeration_time(PCFG, list_algorithms):
# '''
# Retrieve the results and plot
# '''
# for i, (algorithm, algo_name, param) in enumerate(list_algorithms):
# try:
# f = open("results_syntactic/run_2_" + algo_name + '_' + str(seed) + ".pickle","rb")
# run_algo = pickle.load(f)
# f.close()
# except:
# print("algorithm not run yet")
# assert(False)
# processed_run_algo = experiment_enumeration_time(run_algo)
# plt.scatter([x for x in range(1,len(processed_run_algo)+1)], processed_run_algo, label = algo_name, s = 8)
# #min_proba = 0.5
# plt.legend()
# plt.xlabel("number of programs")
# plt.ylabel("time (in seconds)")
# plt.title(title)
# #plt.xscale('log')
# plt.yscale('log')
# #plt.show()
# plt.savefig("results_syntactic/enumeration_time_%s.png" % seed, dpi=300, bbox_inches='tight')
# plt.clf()
# plot_enumeration_time(pcfg, list_algorithms)
####################
# Third experiment #
####################
# probability programs versus search time
# paramaters for the dataset
#Create a dataset, number_samples programs with proba in [1O^(-(i+1),1O^(-i)] for i in [imin, imax]
imin = 2
imax = 10
number_samples = 10
#title = "probability versus search time"
title = ""
# others parameters
total_number_programs = 150_000
timeout = 20 # in seconds
random.seed(seed)
dsl = deepcoder
pcfg = deepcoder_PCFG_t
#list_algorithms = [(heap_search, 'heap search', {'dsl' : dsl, 'environments': {}}),(a_star, 'A*', {})]
#list_algorithms = [(heap_search, 'heap search', {}), (sqrt_sampling, 'SQRT', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.001, 'scale_factor' : 10}), (a_star, 'A*', {})]
list_algorithms = [(heap_search, 'heap search', {}), (sqrt_sampling, 'SQRT', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (a_star, 'A*', {}), (bfs, 'bfs', {'beam_width' : 50000})]
recompute_from_scratch = True
if recompute_from_scratch:
dataset = create_dataset(pcfg)
with open('results_syntactic/dataset_3_'+ str(seed) + ".pickle","wb" ) as f:
pickle.dump(dataset, f)
for algo, algo_name, param in list_algorithms:
run_algo = run_algorithm(dsl, pcfg, algo, param)
with open("results_syntactic/run_3_" + algo_name + '_' + str(seed) + ".pickle","wb" ) as f:
pickle.dump(run_algo, f)
def experiment_probability_vs_time(dataset, run_algo):
result = []
for program,proba in dataset:
hash_program = str(program)
if hash_program in run_algo:
result.append((proba,run_algo[hash_program][1]))
else:
result.append((proba, timeout))
return result
def plot_probability_vs_time(PCFG, list_algorithms):
'''
Retrieve the results and plot
'''
with open('results_syntactic/dataset_3_'+ str(seed) + ".pickle", 'rb') as f:
dataset = pickle.load(f)
for i, (algorithm, algo_name, param) in enumerate(list_algorithms):
try:
f = open("results_syntactic/run_3_" + algo_name + '_' + str(seed) + ".pickle","rb")
run_algo = pickle.load(f)
f.close()
except:
print("algorithm not run yet")
assert(False)
processed_run_algo = experiment_probability_vs_time(dataset, run_algo)
plt.scatter([x for (x,y) in processed_run_algo], [y for (x,y) in processed_run_algo], label = algo_name, s = 8)
#min_proba = 0.5
#plt.xlim((0,min_proba))
plt.legend()
plt.xlabel("probability")
plt.ylabel("search time (in seconds)")
plt.title(title)
plt.xscale('log')
plt.yscale('log')
#plt.show()
plt.savefig("results_syntactic/proba_vs_search_time_%s.png" % seed, dpi=300, bbox_inches='tight')
plt.clf()
plot_probability_vs_time(pcfg, list_algorithms)
#####################
# Fourth experiment #
#####################
# cumulative proba versus number of programs output
# parameters
#title = "cumulative probability versus number of programs output"
title = ""
timeout = 30 # in seconds
random.seed(seed)
total_number_programs = 150_000
dsl = deepcoder
pcfg = deepcoder_PCFG_t
threshold_probability = 1
recompute_from_scratch = False
#list_algorithms = [(heap_search, 'heap search', {'dsl' : dsl, 'environments': {}}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (sqrt_sampling, 'SQRT', {}), (a_star, 'A*', {})]
#list_algorithms = [(heap_search, 'heap search', {}), (sqrt_sampling, 'SQRT', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10})]
list_algorithms = [(heap_search, 'heap search', {}), (sqrt_sampling, 'SQRT', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (bfs, 'bfs', {'beam_width' : 50000})]
if recompute_from_scratch:
for algo, algo_name, param in list_algorithms:
run_algo = run_algorithm(dsl, pcfg, algo, param)
with open("results_syntactic/run_4_" + algo_name + '_' + str(seed) + ".pickle","wb" ) as f:
pickle.dump(run_algo, f)
def experiment_cumulative_vs_trials(run_algo):
result = [(run_algo[e][0], run_algo[e][1], run_algo[e][2]) for e in run_algo] #N, chrono, proba
result.sort(key = lambda x: x[0])
cumulative = 0
for i in range(0, len(result)):
proba = result[i][2]
cumulative+=proba
result[i] = (cumulative, result[i][1])
# result.append((cumulative, chrono))
return result
def plot_cumulative_vs_trials(PCFG, list_algorithms):
'''Retrieve the results and plot'''
for i, (algorithm, algo_name, param) in enumerate(list_algorithms):
try:
f = open("results_syntactic/run_1_" + algo_name + '_' + str(seed) + ".pickle","rb")
run_algo = pickle.load(f)
f.close()
except:
print("algorithm not run yet")
assert(False)
processed_run_algo = experiment_cumulative_vs_trials(run_algo)
plt.scatter([x for (x,y) in processed_run_algo], list(range(1,len(processed_run_algo)+1)), label = algo_name, s = 8)
#plt.xlim((0,threshold_probability))
plt.legend()
plt.xlabel("cumulative probability")
plt.ylabel("number of programs output")
plt.title(title)
# plt.xscale('log')
plt.yscale('log')
#plt.show()
plt.savefig("results_syntactic/cumulative_time_versus_trials_%s.png" % seed, dpi=500, bbox_inches='tight')
plt.clf()
plot_cumulative_vs_trials(pcfg, list_algorithms)
###############################
##########
##
#------------------------------
# TO BE REMOVED
##################
# Create dataset #
##################
# imin = 1
# imax = 11
# timeout = 50 # in seconds
# # total_number_programs = 1_000_000 # 1M programs
# number_samples = 10
# total_number_programs = 1_000_000 #10_000_000 # 1M programs
# # # PCFG
# # deepcoder = DSL(semantics, primitive_types, no_repetitions)
# # t = Arrow(List(INT),List(INT))
# # deepcoder_CFG_t = deepcoder.DSL_to_CFG(t)
# # deepcoder_PCFG_t = deepcoder.DSL_to_Uniform_PCFG(t)
# # deepcoder_PCFG_t.put_random_weights(alpha = .7)
# # DATASET = create_dataset(deepcoder_PCFG_t)
# # print(DATASET)
# # Set of algorithms to test
# list_algorithms = [(heap_search, 'heap search', {}), (dfs, 'dfs', {}), (threshold_search, 'threshold', {'initial_threshold' : 0.0001, 'scale_factor' : 10}), (sqrt_sampling, 'SQRT', {}), (a_star, 'A*', {})]
# # (bfs, 'bfs', {"beam_width" : 5000})
# # Set of PCFG
# #PCFG_test = [(deepcoder, deepcoder_PCFG_t)] # (dsl, PCFG)
# # First experiment
# # def plot_cumulative_vs_time(PCFG, list_algorithms, list_result):
# # '''
# # Retrieve the results and plot
# # '''
# # min_proba = 1
# # for i, (algorithm, name, param) in enumerate(list_algorithms):
# # # if param == []:
# # # with open('experiment_results/proba_vs_search_time_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__), 'rb') as f:
# # # result = pickle.load(f)
# # # else:
# # # with open('experiment_results/proba_vs_search_time_%s_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__, param[0]), 'rb') as f:
# # # result = pickle.load(f)
# # result = list_results[i]
# # min_proba = min(min_proba,result[-1][0])
# # #plt.scatter([x for (x,y) in result], [i for i in range(len(result))], label = name, s = 8)
# # plt.scatter([x for (x,y) in result], [y for (x,y) in result], label = name, s = 8)
# # #min_proba = 0.5
# # plt.xlim((0,min_proba))
# # plt.legend()
# # plt.xlabel("cumulative probability")
# # plt.ylabel("search time (in seconds)")
# # plt.title('PCFG')
# # # plt.xscale('log')
# # plt.yscale('log')
# # plt.show()
# # plt.savefig("images/proba_vs_search_time_%s.png" % 'PCFG', dpi=300, bbox_inches='tight')
# # plt.clf()
# # Second experiment: log search time versus log proba
# def experiment_probability_vs_time(dataset, run_algo):
# result = []
# for program,proba in dataset:
# hash_program = str(program)
# if hash_program in run_algo:
# result.append((proba,run_algo[hash_program][1]))
# else:
# result.append((proba, timeout))
# return result
# def plot_probability_vs_time(PCFG, list_algorithms, list_result):
# '''
# Retrieve the results and plot
# '''
# min_proba = 1
# for i, (algorithm, name, param) in enumerate(list_algorithms):
# # if param == []:
# # with open('experiment_results/proba_vs_search_time_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__), 'rb') as f:
# # result = pickle.load(f)
# # else:
# # with open('experiment_results/proba_vs_search_time_%s_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__, param[0]), 'rb') as f:
# # result = pickle.load(f)
# result = list_results[i]
# min_proba = min(min_proba,result[-1][0])
# #plt.scatter([x for (x,y) in result], [i for i in range(len(result))], label = name, s = 8)
# plt.scatter([x for (x,y) in result], [y for (x,y) in result], label = name, s = 8)
# #min_proba = 0.5
# #plt.xlim((0,min_proba))
# plt.legend()
# plt.xlabel("probability")
# plt.ylabel("search time (in seconds)")
# plt.title('PCFG')
# plt.xscale('log')
# plt.yscale('log')
# plt.show()
# plt.savefig("images/proba_vs_search_time_%s.png" % 'PCFG', dpi=300, bbox_inches='tight')
# plt.clf()
# # Third experiment: enumeration time
# def experiment_enumeration_time(run_algo):
# result = [run_algo[e][1] for e in run_algo] #N, chrono, proba
# result.sort()
# return result
# def plot_enumeration_time(PCFG, list_algorithms, list_result):
# '''
# Retrieve the results and plot
# '''
# for i, (algorithm, name, param) in enumerate(list_algorithms):
# # if param == []:
# # with open('experiment_results/proba_vs_search_time_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__), 'rb') as f:
# # result = pickle.load(f)
# # else:
# # with open('experiment_results/proba_vs_search_time_%s_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__, param[0]), 'rb') as f:
# # result = pickle.load(f)
# result = list_results[i]
# #plt.scatter([x for (x,y) in result], [i for i in range(len(result))], label = name, s = 8)
# plt.scatter([x for x in range(1,len(result)+1)], result, label = name, s = 8)
# #min_proba = 0.5
# #plt.xlim((0,min_proba))
# plt.legend()
# plt.xlabel("number of programs")
# plt.ylabel("time (in seconds)")
# plt.title('PCFG')
# #plt.xscale('log')
# plt.yscale('log')
# plt.show()
# plt.savefig("images/proba_vs_search_time_%s.png" % 'PCFG', dpi=300, bbox_inches='tight')
# plt.clf()
# list_algorithms_2 = [(heap_search, 'heap search', {}),(a_star, 'A*', {})]
# list_results = []
# for algo, name, param in list_algorithms_2:
# globals()[name + "_search_run"] = run_algorithm(deepcoder, deepcoder_PCFG_t, algo, param)
# # with open('experiment_results/proba_vs_search_time_%s_%s_%s.bin' % (G_name, alpha, algorithm.__name__), 'wb') as f:
# # pickle.dump(name + "_search_run", f)
# globals()[name + "_search_run_2"] = experiment_enumeration_time(globals()[name + "_search_run"])
# list_results.append(globals()[name + "_search_run_2"])
# # print(globals()[name + "_search_run_2"][-1])
# #print(list_results[0])
# plot_enumeration_time(deepcoder_PCFG_t, list_algorithms_2, list_results)
#print(heap_search_run)
#dfs_search_run = run_algorithm(deepcoder, deepcoder_PCFG_t, dfs, [])
# first = experiment_cumulative_vs_time(heap_search_run)
# print(first[-1])
# second = experiment_cumulative_vs_time(dfs_search_run)
# print(second[-1])