def net_based_merge(n1, n2, distance, gamma):
# calculates net over point nets and use it as keys of the merged network (similar/same? to the key-based)
nn = create_aggregating_net(gamma=gamma,
apns=list(n1.keys())+list(n2.keys()),
distance=distance)
targ = defaultdict(list)
for k, v in nn.items():
targ[k] = n1.get(k, []).copy()
targ[k] += n2.get(k, []).copy()
for el in v:
targ[k] += n1.get(el, []).copy()
targ[k] += n2.get(el, []).copy()
targ[k].append(el)
return targ
def net_based_multi_merge(nets, distance, gamma):
apns = []
for net in nets:
apns += list(net.keys())
nn = create_aggregating_net(gamma=gamma,
apns=apns,
distance=distance)
targ = defaultdict(list)
for k, v in nn.items():
for net in nets:
targ[k] += net.get(k, []).copy()
for el in v:
for net in nets:
targ[k] += net.get(el, []).copy()
targ[k].append(el)
return targ
def merge_voting_nets(nets, distance, gamma):
apns = []
for net in nets:
apns += list(net.keys())
nn = create_aggregating_net(gamma=gamma,
apns=apns,
distance=distance)
# transfer the "votes" from original networks to just created new anchors
targ = dict()
for k, v in nn.items():
for net in nets:
if k in net:
targ[k] = net.get(k)
break
for el in v:
for net in nets:
if el in net:
targ[k] = list(np.add(targ[k], net[el]))
break
return targ
def single_net(x, gamma):
start = time.time()
net = create_aggregating_net(gamma=gamma, apns=x.keys(), distance=lambda x1, y1: distance(x1, y1, x))
end = time.time()
print(f"Creating network...{gamma=} {len(x)} Elapsed: {end-start}")
return dict(net)
d = lambda x, y: distance(x, y, funcs)
sample_size = 200
smp = mysample(v, sample_size)
funcs = smp.groupby(by='apn')['nf'].apply(set)
#print('reading test ste')
#tests = pd.read_csv('res2/9500-test.csv', index_col=0)
test_size = 20
train, test = train_test_split(funcs, test_size=test_size, random_state=42)
test.to_csv(f"res/test-{test_size}.csv")
#for gamma in tqdm([0, 1, 2, 4, 8, 16, 32, 180, 192]):
#for gamma in tqdm([0, 0.1, 0.4, 0.5, 0.7, 0.8, 0.85, 0.9, 1.0]):
intervals = 18
for gamma in tqdm([x * 1/intervals for x in range(0, intervals+1)]):
merged = generate_merged(gamma=gamma, funcs=funcs)
#onsm, nets = generate_merged(gamma=gamma, distance=distance, labels=labels)
print("Creating reference netwrok")
start = time.time()
reference = create_aggregating_net(gamma=gamma, apns=train.index, distance=d)
end = time.time()
print(f"\tElapsed: {end-start}")
with open(f"res/mergers-jaccard-{gamma}.pickle", 'wb+') as f:
pickle.dump([dict(reference), dict(merged)], f)