def get_sway_res(model):
# generating the 10k random solutions
candidates = list()
for _ in range(10000):
ran_dec = [random.random() for _ in range(model.decsNum)]
can = model.Individual(ran_dec)
candidates.append(can)
global M
M = model
res = sway(candidates, model.eval, where, comparing)
return res
def get_sway_res(model):
# load the 10k sat solutions
# with open('./tse_rs/' + model.name + '.txt', 'r') as f:
# candidates = list()
# for l in f:
# can = model.Individual(l.strip('\n'))
# candidates.append(can)
candidates = sat_gen_valid_pop(model, 10000)
res = sway(candidates, model.eval,
partial(split_products, groupC=min(15, model.featureNum // 7)),
comparing)
return res
def get_sway_res(model):
def _dist(ind1, ind2):
d = 0
for i, j in zip(ind1, ind2):
d += (i - j)**2
return d
def _where(pop):
rand = random.choice(pop)
ds = [_dist(i, rand) for i in pop]
east = pop[ds.index(max(ds))]
ds = [_dist(i, east) for i in pop]
west = pop[ds.index(max(ds))]
c = _dist(east, west)
cc = 2 * c**0.5
mappings = list()
for x in pop:
a = _dist(x, west)
b = _dist(x, east)
d = (a + c - b) / cc
mappings.append((x, d))
mappings = sorted(mappings, key=lambda i: i[1])
mappings = [i[0] for i in mappings]
n = len(mappings)
eastItems = mappings[:int(n * 0.2)] + mappings[int(n * 0.5):int(n *
0.8)]
westItems = mappings[int(n * 0.2):int(n * 0.5)] + mappings[int(n *
0.8):]
# westItems = mappings[len(mappings)//2:]
return west, east, eastItems, westItems
def _comparing(part1, part2):
return bin_dominate(part1, part2)
# Starting SWAY execution here
model.eval_counts = 0
candidates = list()
for _ in range(10000):
ran_dec = [random.random() for _ in range(model.decsNum)]
can = model.Individual(ran_dec)
candidates.append(can)
res = sway(candidates, functools.partial(model.eval, normalized=False),
_where, _comparing)
return emo.sortNondominated(res, len(res), first_front_only=True)[0]
def tcp_sway(dataset, suite, ver, initial, stop, alg=2):
# num_tests_dict = {'flex':21, 'grep':193, 'gzip':211, 'sed':36}
num_tests_dict = {
'flex': {
1: 21,
2: 525
},
'grep': {
1: 193,
2: 152,
3: 140
},
'gzip': {
1: 211
},
'sed': {
1: 36,
2: 360
}
}
# Compare function
def comparing(part1, part2): ##Need modification
if ver == '0':
return get_apsd(dataset, part1) > get_apsd(dataset, part2)
else:
return get_apsd_linux(dataset, suite, ver, part1) > get_apsd_linux(
dataset, suite, ver, part2)
if ver == '0':
path = 'Datasets/' + dataset + '/traces'
file_list = os.listdir(path)
length = sum(['dump' in name for name in file_list])
else: # linux utils
length = num_tests_dict[dataset][suite]
## Build initial population in a parallel manner
seed = np.array(range(1, length + 1))
wrapper = partial(permute_row, seed)
pool = Pool()
new_rows = pool.map(wrapper, range(1, initial))
pool.close()
candidates = np.concatenate((np.reshape(seed,
(1, length)), np.array(new_rows)))
# candidates = []
# for _ in range(initial):
# x = list(range(1, length + 1))
# while x in candidates: # avoid any repetitions
# random.shuffle(x)
# candidates.append(x)
## Binary embedding
if alg == 1:
raise ValueError("Binary embedding not yet implemented!!")
# emb_cand, emb_dict = embed(candidates)
# res = sway(emb_cand, partial(split_products, groupC=min(15, dim // 7)), comparing, alg, emb_dict)
# return res, candidates
## Continuous embedding
elif alg == 2:
return sway(candidates, where, comparing, stop, alg, None), candidates
def optimize(init_pop, eval_func):
import warnings
warnings.filterwarnings('ignore')
return sway(init_pop, eval_func, split_products, comparing)
def optimize(init_pop, eval_func):
import warnings
warnings.filterwarnings('ignore')
res = sway(init_pop, eval_func, where, comparing)
return res
