-
Notifications
You must be signed in to change notification settings - Fork 0
/
lab-4-mutations.py
370 lines (316 loc) · 12.8 KB
/
lab-4-mutations.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
from collections import defaultdict
import itertools
from itertools import groupby
from typing import Iterable, Generator
import matplotlib.pylab as plt
import numpy as np
import seaborn as sns
from figures import ensure_directory
import src.resources as resources
from src.resources.models import SaveRecord
def read_name(record: SaveRecord) -> str:
return record.meta.arguments['genformat']
def groupby_representation(records: list[SaveRecord]) -> dict[str, list[SaveRecord]]: return {
representation: list(values)
for (representation, values) in groupby(records, read_name)
}
# def aggregated(records: list[SaveRecord]):
# plt.tight_layout()
# plt.figure(figsize=(12, 5))
#
# generations = len(records[0].history)
# representations = {
# representation: (aggregate(values, np.mean), aggregate(values, np.std))
# for representation, values in groupby_representation(records).items()
# }
#
# ticks = range(1, generations + 1)
# for (representation, (averages, stddevs)) in representations.items():
# plt.plot(ticks, averages, label=f"f{representation}")
# stddevs = [
# stddev / 5
# for stddev in stddevs
# ]
#
# plt.fill_between(
# ticks,
# dif(averages, stddevs),
# sum(averages, stddevs),
# alpha=0.3
# )
#
# plt.ylabel('Vertpos')
# plt.legend(
# loc='center left',
# bbox_to_anchor=(0.96, 0.5),
# )
# plt.savefig(f'resources/lab-4/figures/1/all-aggregated.png', bbox_inches='tight')
def within_constraint(genotype, values, criterion, max_value):
REPORT_CONSTRAINT_VIOLATIONS = True
if max_value is None: return True
actual_value = values[criterion]
if actual_value > max_value:
if REPORT_CONSTRAINT_VIOLATIONS:
print(
f'Genotype "{genotype}" assigned low fitness because it violates constraint "{criterion}": {actual_value} exceeds threshold {max_value}'
)
return False
return True
def frams_evaluate(lib, individual):
unfit = [-1] * 1
genotype = individual[0]
valid = True
try:
evaluation = lib.evaluate([genotype])[0]['evaluations'][""]
fitness = [evaluation["vertpos"]]
# fitness = [
# evaluation[target]
# if evaluation[target] > 0 else
# evaluation[target] + (evaluation['numparts'] / constants.max_numparts / 5)
# for target in OptimizationTargets
# ]
evaluation['numgenocharacters'] = len(genotype)
valid &= within_constraint(genotype, evaluation, 'numparts', 15)
valid &= within_constraint(genotype, evaluation, 'numjoints', 30)
valid &= within_constraint(genotype, evaluation, 'numneurons', 20)
valid &= within_constraint(genotype, evaluation, 'numconnections', 30)
valid &= within_constraint(genotype, evaluation, 'numgenocharacters', None)
if not valid: return unfit
except (KeyError, TypeError) as error:
# the evaluation may have failed for an invalid genotype
# (such as X[@][@] with "Don't simulate genotypes with warnings" option) or for some other reason.
print(
f'Problem "{error}" so could not evaluate genotype "{genotype}", hence assigned it low fitness: {unfit}'
)
return unfit
return fitness
def generate_mutations(lib, individual, count: int):
mutations = []
scores = []
while len(mutations) < count:
mutated = lib.mutate(individual)
[mutation] = mutated
if mutation in mutations: continue
[score] = frams_evaluate(lib, mutated)
if score < 0: continue
mutations.append(mutation)
scores.append(score)
return mutations, scores
def generate_crossovers(lib, first, second, count: int):
crossovers = []
scores = []
iters = 0
while len(crossovers) < count:
iters += 1
if iters > 100: break
cross = lib.crossOver(first, second)
if cross in crossovers: continue
[score] = frams_evaluate(lib, [cross])
if score < 0: continue
crossovers.append(cross)
scores.append(score)
return crossovers, scores
import os
import pickle
from functools import wraps
Individual = tuple[str, float]
def cache_pickle(path: str, name: str = 'data'):
def decorator(fn):
@wraps(fn)
def wrapper(*args, **kwargs):
if os.path.exists(path):
print(f"Loaded {name} from {path}...")
with open(path, 'rb') as file:
return pickle.load(file)
cache = fn(*args, **kwargs)
print(f"Stored {name} in {path}...")
with open(path, 'wb') as file:
pickle.dump(cache, file)
return cache
return wrapper
return decorator
@cache_pickle('.lab-4-pickle-records', 'records')
def read_records():
return [resources.read(name, model=SaveRecord) for name in resources.names("./lab-4/results")]
def prune_records(records: list[SaveRecord], threshold: float):
return [record for record in records if record.population[0].values["vertpos"] > threshold]
@cache_pickle('.lab-4-pickle-mutations', 'mutations')
def create_mutations(records: list[SaveRecord]) -> dict[
str, list[tuple[tuple[str, float], tuple[list[str], list[float]]]]
]:
from libs.framspy.FramsticksLib import FramsticksLib
lib = FramsticksLib(*map(records[0].meta.arguments.get, ('path', 'lib', 'sim')))
processed = {}
groups = groupby_representation(records)
for (representation, group) in groups.items():
items = []
for (i, record) in enumerate(group):
print(f'creating mutations for {representation} | {i + 1}/{len(group)}')
genotype = record.population[0].genotype
score = record.population[0].values["vertpos"]
items.append(((genotype, score), generate_mutations(lib, [genotype], 32)))
processed[representation] = items
return processed
@cache_pickle('.lab-4-pickle-crossovers', 'crossovers')
def create_crossovers(records: list[SaveRecord]) -> dict[
str, list[tuple[tuple[Individual, Individual], tuple[list[str], list[float]]]]
]:
from libs.framspy.FramsticksLib import FramsticksLib
lib = FramsticksLib(*map(records[0].meta.arguments.get, ('path', 'lib', 'sim')))
processed = {}
groups = groupby_representation(records)
for (representation, group) in groups.items():
group = group[:100]
items: list[tuple[tuple[tuple[str, float], tuple[str, float]], tuple[list[str], list[float]]]] = []
combination_count = len(group) * (len(group) - 1) // 2
for (i, (record_a, record_b)) in enumerate(itertools.combinations(group, r=2)):
print(f'creating crossovers for {representation} | {i + 1}/{combination_count}')
genotype_a = record_a.population[0].genotype
score_a = record_a.population[0].values["vertpos"]
genotype_b = record_b.population[0].genotype
score_b = record_b.population[0].values["vertpos"]
crossovers = generate_crossovers(lib, genotype_a, genotype_b, 2)
if not crossovers: continue
items.append((((genotype_a, score_a), (genotype_b, score_b)), crossovers))
processed[representation] = items
return processed
@cache_pickle('.lab-4-pickle-random-walk', 'random-walk')
def random_walks(records: Iterable[SaveRecord]):
from libs.framspy.FramsticksLib import FramsticksLib
lib = FramsticksLib(*map(records[0].meta.arguments.get, ('path', 'lib', 'sim')))
def create_population(records: list[SaveRecord]) -> list[Individual]: return [
(record.population[0].genotype, record.population[0].values["vertpos"])
for record in records
]
def create_buckets(population: Iterable[Individual], count: int):
min_score = min(score for _, score in population)
max_score = max(score for _, score in population)
print(min_score, max_score)
buckets = {
(a, b): []
for (a, b) in itertools.pairwise(np.linspace(min_score, max_score + 0.05, count + 1))
}
for (genotype, score) in population:
for (lower, upper), bucket in buckets.items():
if lower <= score < upper:
bucket.append(genotype)
break
return buckets
def random_walk(start: str, iterations: int) -> Generator[Individual, None, None]:
origin = [start]
[score] = frams_evaluate(lib, origin)
yield origin, score
while iterations > 0:
mutated = lib.mutate(origin)
[score] = frams_evaluate(lib, mutated)
if score < 0: continue
iterations -= 1
origin = mutated
yield origin, score
processed = {}
for (representation, group) in groupby_representation(records).items():
population = create_population(group)
buckets = create_buckets(population, 5)
items = defaultdict(list)
for (lower, upper), bucket in buckets.items():
for (i, genotype) in enumerate(bucket):
print(f'creating random walks for {representation} | {lower:.2f} - {upper:.2f} | {i + 1}/{len(bucket)}')
walk = tuple(random_walk(genotype, 25))
items[(lower, upper)].append(walk)
processed[representation] = items
return processed
def main():
records = read_records()
for (representation, items) in groupby_representation(records).items():
print(f"before {representation} prunning: {len(items):>5}")
records = prune_records(records, 0.05)
for (representation, items) in groupby_representation(records).items():
print(f"after {representation} prunning: {len(items):>5}")
# boxplot(records)
# aggregated(records)
mutations = create_mutations(records)
crossovers = create_crossovers(records)
walks = random_walks(records)
ensure_directory(f'resources/lab-4/figures/1')
def scatterplots(mutations: dict[str, list[tuple[str, list[float]]]]):
for (representation, group) in mutations.items():
points_x = []
points_y = []
for ((original, score), (mutations, scores)) in group:
y = [score] + scores
x = [score] * len(y)
points_x.extend(x)
points_y.extend(y)
plt.tight_layout()
plt.figure(figsize=(12, 5))
plt.scatter(points_x, points_y, alpha=0.1, color='green')
plt.title(f"Representation f{representation}")
plt.ylabel("Mutation Fitness")
plt.xlabel("Original Fitness")
plt.savefig(f'resources/lab-4/figures/1/mutations-{representation}.png', bbox_inches='tight')
# plt.show()
def heatmaps(crossovers):
plt.tight_layout()
plt.figure(figsize=(12, 5))
for (representation, group) in crossovers.items():
parent_fitness_of = (
{a: score for ((a, score), _), _ in group}
|
{b: score for (_, (b, score)), _ in group}
)
child_fitness_of = (
{(a, b): score for ((a, _), (b, _)), (_, [score, *_]) in group}
)
parents_a = sorted([a for ((a, _), _), _ in group], key=lambda a: parent_fitness_of[a])
parents_b = sorted([b for (_, (b, _)), _ in group], key=lambda b: parent_fitness_of[b])
points = np.zeros((len(parents_b), len(parents_a)))
n = len(parents_a)
for (i, b) in enumerate(parents_b):
for (j, a) in enumerate(parents_a):
points[n - i - 1, j] = child_fitness_of.get((a, b)) or 0
sns.heatmap(points)
plt.title(f"Representation f{representation}")
plt.xlabel("First Parent Fitness")
plt.ylabel("Second Parent Fitness")
plt.xticks([0, (len(points) - 1) // 2, len(points) - 1], ['0', '1', f'{parent_fitness_of[parents_b[-1]]:.2f}'], rotation=0)
plt.yticks([0, (len(points) - 1) // 2, len(points) - 1], [f"{parent_fitness_of[parents_a[-1]]:.2f}", '1', '0'], rotation=0)
plt.savefig(f'resources/lab-4/figures/1/crossovers-{representation}.png', bbox_inches='tight')
plt.show()
def walkplots(walks):
for (representation, buckets) in walks.items():
plt.tight_layout()
plt.figure(figsize=(12, 5))
legends = [f'Range: [{a:0.2f}-{b:0.2f})' for (a, b) in buckets]
legends[-1] = legends[-1].replace(')', ']')
plt.title(f'Representation f{representation}')
plt.xlabel('Iteration')
plt.ylabel('Fitness')
ticks = range(0, 26)
buckets_scores = {ranges: [[score for (_, score) in run] for run in runs] for (ranges, runs) in buckets.items()}
for i, (ranges, runs) in enumerate(buckets_scores.items()):
# print(len(runs))
averages = [np.mean([run[iteration] for run in runs]) for iteration in ticks]
stddevs = [np.std([run[iteration] for run in runs]) for iteration in ticks]
stddevs = [stddev / 5 for stddev in stddevs]
print(averages, stddevs)
def addlists(a, b): return [x + y for (x, y) in zip(a, b)]
def sublists(a, b): return [x - y for (x, y) in zip(a, b)]
plt.plot(ticks, averages, label=legends[i])
plt.fill_between(
ticks,
addlists(averages, stddevs),
sublists(averages, stddevs),
alpha=0.3
)
# plt.savefig(f'resources/lab-4/figures/1/walks-{representation}.png', bbox_inches='tight')
plt.legend(
loc='center left',
bbox_to_anchor=(0.96, 0.5),
)
plt.savefig(f'resources/lab-4/figures/1/walks-{representation}.png', bbox_inches='tight')
plt.show()
# scatterplots(mutations)
# heatmaps(crossovers)
walkplots(walks)
if __name__ == '__main__':
main()